Soros is a compound AI system built carefully from the ground up to trace a path (multiple paths, really) from a description of a geopolitical event all the way to capital market implications.
* Here's how we set it up:
Given a description of a given geopolitical event (can be a couple of words; the demo literally has "US-Iran conflict" as the entire string), Soros will - \n(1) first analyze and perform deep research on it, running scores of searches in parallel to gather deep context that's time-weighted for real events and can serve as background for hypothetical events ("PRC-Taiwan reunification crisis 2027") \n(2) map out relevant individual actors, factions, organizations, and their propensities, capabilities and salience under a variety of sociopolitical, military, and socioeconomic axes,\n(3) determine the key resources (or geopolitical chokepoints) whose control is being "negotiated" or fought over,\n(4) identify the landscape of key decisions that a subset of actors need to take, and the constraints and strategic options they have for each one,\n(5) generate forward-looking scenarios that incorporate potential paths weaving through each of the key decisions, \n(6) engage a full-blown Monte Carlo simulation engine and generate thousands of trajectories to estimate relative probabilities of each of the scenarios coming to pass,\n(7) analyze each scenario to generate likely capital flows and identify the sectors, industries, companies, currencies, and commodities most affected (direction and horizon)\n(8) identify key search phrases and X/Twitter accounts to track in order to periodically update the analysis
This is obviously a fairly complicated pipeline, with lots of moving components and potential failure points. In order to mitigate the worst aspects of this, we engage the services of Pyrrho (yup, we named it after the Greek philosopher dude), an AI agent that we have set up to be the harshest possible critic of Soros' intermediate and final outputs. Each step above is a delicate dance between Soros and Pyrrho, and this interaction serves to enhance the quality of the final output dramatically.
Once you have the analysis setup, you can perform the now-classic "Chat with Analysis" interaction by using the "Ask Soros" functionality. We have a separate chat model hooked up that is (hopefully sufficiently) guard-railed and context-injected enough to focus completely and exclusively on answering freeform questions about the analysis.
In the live non-demo system, the user has multiple ways of engaging further with the analysis: they can add new (private) information and do a re-run, they can mark out specific items from associated X/Twitter/search feeds, they can add new actors and resources, modify existing ones, delete some as needed, and basically run simulation after simulation to test out hypotheses (e.g. "What if China entered the conflict? What if France sent its nuclear subs to patrol the Straits of Hormuz?" etc.).
You can see the results of all of this, and more, at www.asksoros.com - there is a statically-served demo analysis of the current US-Iran conflict; we urge you to "Take a tour" of the interface to familiarize yourselves with it.
(Continuing the post with the first comment below..)"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"Show HN: Soros \u2013 AI for geopolitical macro investing"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.asksoros.com"}},"_tags":["story","author_muggermuch","story_47418553","show_hn"],"author":"muggermuch","children":[47418557,47419315,47420894,47422768],"created_at":"2026-03-17T21:26:26Z","created_at_i":1773782786,"num_comments":10,"objectID":"47418553","points":26,"story_id":47418553,"story_text":"Hi HN! We are Anshuman and Kar\u00e9n, the co-founders of Lookback Labs and the co-designers of Soros (https://www.asksoros.com/).
Soros is a compound AI system built carefully from the ground up to trace a path (multiple paths, really) from a description of a geopolitical event all the way to capital market implications.
* Here's how we set it up:
Given a description of a given geopolitical event (can be a couple of words; the demo literally has "US-Iran conflict" as the entire string), Soros will - \n(1) first analyze and perform deep research on it, running scores of searches in parallel to gather deep context that's time-weighted for real events and can serve as background for hypothetical events ("PRC-Taiwan reunification crisis 2027") \n(2) map out relevant individual actors, factions, organizations, and their propensities, capabilities and salience under a variety of sociopolitical, military, and socioeconomic axes,\n(3) determine the key resources (or geopolitical chokepoints) whose control is being "negotiated" or fought over,\n(4) identify the landscape of key decisions that a subset of actors need to take, and the constraints and strategic options they have for each one,\n(5) generate forward-looking scenarios that incorporate potential paths weaving through each of the key decisions, \n(6) engage a full-blown Monte Carlo simulation engine and generate thousands of trajectories to estimate relative probabilities of each of the scenarios coming to pass,\n(7) analyze each scenario to generate likely capital flows and identify the sectors, industries, companies, currencies, and commodities most affected (direction and horizon)\n(8) identify key search phrases and X/Twitter accounts to track in order to periodically update the analysis
This is obviously a fairly complicated pipeline, with lots of moving components and potential failure points. In order to mitigate the worst aspects of this, we engage the services of Pyrrho (yup, we named it after the Greek philosopher dude), an AI agent that we have set up to be the harshest possible critic of Soros' intermediate and final outputs. Each step above is a delicate dance between Soros and Pyrrho, and this interaction serves to enhance the quality of the final output dramatically.
Once you have the analysis setup, you can perform the now-classic "Chat with Analysis" interaction by using the "Ask Soros" functionality. We have a separate chat model hooked up that is (hopefully sufficiently) guard-railed and context-injected enough to focus completely and exclusively on answering freeform questions about the analysis.
In the live non-demo system, the user has multiple ways of engaging further with the analysis: they can add new (private) information and do a re-run, they can mark out specific items from associated X/Twitter/search feeds, they can add new actors and resources, modify existing ones, delete some as needed, and basically run simulation after simulation to test out hypotheses (e.g. "What if China entered the conflict? What if France sent its nuclear subs to patrol the Straits of Hormuz?" etc.).
You can see the results of all of this, and more, at www.asksoros.com - there is a statically-served demo analysis of the current US-Iran conflict; we urge you to "Take a tour" of the interface to familiarize yourselves with it.
(Continuing the post with the first comment below..)","title":"Show HN: Soros \u2013 AI for geopolitical macro investing","updated_at":"2026-04-07T21:50:26Z","url":"https://www.asksoros.com"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"mgh2"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"Journalist Karen Hao on Sam Altman, OpenAI and the \"Quasi-Religious\" Push for AI [video]"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.youtube.com/watch?v=s4hZz9Vd0lY"}},"_tags":["story","author_mgh2","story_44631214"],"author":"mgh2","created_at":"2025-07-21T02:19:14Z","created_at_i":1753064354,"num_comments":0,"objectID":"44631214","points":4,"story_id":44631214,"title":"Journalist Karen Hao on Sam Altman, OpenAI and the \"Quasi-Religious\" Push for AI [video]","updated_at":"2025-07-21T07:27:04Z","url":"https://www.youtube.com/watch?v=s4hZz9Vd0lY"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"tikkun"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["karen"],"value":"Inflection \u2013 Co-Founded by Mustafa Suleyman, Reid Hoffman and Kar\u00e9n Simonyan"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"https://inflection.ai/about/"}},"_tags":["story","author_tikkun","story_34642962"],"author":"tikkun","children":[34642963],"created_at":"2023-02-03T16:06:43Z","created_at_i":1675440403,"num_comments":1,"objectID":"34642962","points":2,"story_id":34642962,"title":"Inflection \u2013 Co-Founded by Mustafa Suleyman, Reid Hoffman and Kar\u00e9n Simonyan","updated_at":"2024-09-20T13:14:23Z","url":"https://inflection.ai/about/"},{"_highlightResult":{"author":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"rigid_airship"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["karen"],"value":"Bye Karen"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["karen"],"value":"https://blog.iambob.me/bye-karen/"}},"_tags":["story","author_rigid_airship","story_29744256"],"author":"rigid_airship","created_at":"2021-12-31T02:38:01Z","created_at_i":1640918281,"num_comments":0,"objectID":"29744256","points":2,"story_id":29744256,"title":"Bye Karen","updated_at":"2024-09-20T10:08:36Z","url":"https://blog.iambob.me/bye-karen/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"techinterests"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"I built PowerShellGPT, a native Windows app that connects large language models (ChatGPT, Claude, and even local models like LM Studio) directly to PowerShell and a live embedded browser, enabling full voice or text-controlled automation of both your system and the web.
Unlike traditional chat interfaces, PowerShellGPT doesn\u2019t just generate code\u2014it executes it, watches the results, and loops that feedback back to the model for error correction or follow-up actions. It acts more like an autonomous agent than a passive chatbot.
Core Features:\nNatural Language to Execution Loop: AI generates PowerShell or JavaScript, executes it, sees the result, and corrects itself if needed.
Voice Recognition: Use voice commands in over 80 languages to control your system or web browser.
@PowerShellGPT@ and @JsGPT@ Tags: Triggers live execution in embedded PowerShell and browser environments.
Agent Bridge: Enables multiple named AI agents to communicate and delegate tasks to each other.
Keyword Commands: Create templates like \u201cemail [KEYWORD] to Karen\u201d or \u201csearch Google for [KEYWORD]\u201d and fill them in with your voice.
Chain & Wait Commands: Link multiple actions together and add delays between them like:\nsearch YouTube for music and then wait 5 seconds and then play the first result
Voice Activated Automation: Detect faces, trigger system events, play media, fill out forms\u2014entirely hands-free.
Cross-language Support: Generate and run Python, Node.js, Ruby, or C# files using natural language via a system I call MACFARI ("Make A Code File And Run It").
Use Cases:\nSystem automation and scripting with no typing
Web task automation with dynamic JavaScript injection
Creating AI-powered assistants that collaborate via voice
Accessibility-friendly computing via full voice control
The app runs as a lightweight Delphi GUI with two embedded browsers: one for AI, one for execution. All AI-to-execution communication happens via tag parsing and a secure permission system. You can toggle access prompts, enable/disable looping protection, and modify wake word settings.
https://youtube.com/@PowerShellGPT
Would love feedback, ideas, or feature requests. I'm especially curious what you\u2019d build with it\u2014or how you'd break it.
Thanks for checking it out!"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"Show HN: PowerShellGPT Voice Controlled AI Agent Automates Windows and the Web"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://powershellgpt.com/"}},"_tags":["story","author_techinterests","story_43729306","show_hn"],"author":"techinterests","created_at":"2025-04-18T16:00:54Z","created_at_i":1744992054,"num_comments":0,"objectID":"43729306","points":1,"story_id":43729306,"story_text":"I built PowerShellGPT, a native Windows app that connects large language models (ChatGPT, Claude, and even local models like LM Studio) directly to PowerShell and a live embedded browser, enabling full voice or text-controlled automation of both your system and the web.
Unlike traditional chat interfaces, PowerShellGPT doesn\u2019t just generate code\u2014it executes it, watches the results, and loops that feedback back to the model for error correction or follow-up actions. It acts more like an autonomous agent than a passive chatbot.
Core Features:\nNatural Language to Execution Loop: AI generates PowerShell or JavaScript, executes it, sees the result, and corrects itself if needed.
Voice Recognition: Use voice commands in over 80 languages to control your system or web browser.
@PowerShellGPT@ and @JsGPT@ Tags: Triggers live execution in embedded PowerShell and browser environments.
Agent Bridge: Enables multiple named AI agents to communicate and delegate tasks to each other.
Keyword Commands: Create templates like \u201cemail [KEYWORD] to Karen\u201d or \u201csearch Google for [KEYWORD]\u201d and fill them in with your voice.
Chain & Wait Commands: Link multiple actions together and add delays between them like:\nsearch YouTube for music and then wait 5 seconds and then play the first result
Voice Activated Automation: Detect faces, trigger system events, play media, fill out forms\u2014entirely hands-free.
Cross-language Support: Generate and run Python, Node.js, Ruby, or C# files using natural language via a system I call MACFARI ("Make A Code File And Run It").
Use Cases:\nSystem automation and scripting with no typing
Web task automation with dynamic JavaScript injection
Creating AI-powered assistants that collaborate via voice
Accessibility-friendly computing via full voice control
The app runs as a lightweight Delphi GUI with two embedded browsers: one for AI, one for execution. All AI-to-execution communication happens via tag parsing and a secure permission system. You can toggle access prompts, enable/disable looping protection, and modify wake word settings.
https://youtube.com/@PowerShellGPT
Would love feedback, ideas, or feature requests. I'm especially curious what you\u2019d build with it\u2014or how you'd break it.
Thanks for checking it out!","title":"Show HN: PowerShellGPT Voice Controlled AI Agent Automates Windows and the Web","updated_at":"2025-04-18T16:03:47Z","url":"https://powershellgpt.com/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"arowe"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"Built this solo over the past year. You build a deck from 100+ cards - Karen-zilla, Nigerian Scamurai, Dumpster Fire Angel, Florida Man - and fight AI opponents across a 3-episode campaign.
Can't die from low player count. Free account required to play.
Stack: Laravel + Vue 3 + TypeScript.
Would love feedback on whether the first game makes sense without hand-holding."},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Faction Fiasco \u2013 single-player card battler, 100 unhinged characters"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://factionfiasco.com/"}},"_tags":["story","author_arowe","story_48027794","show_hn"],"author":"arowe","children":[48027829,48028008],"created_at":"2026-05-05T20:08:20Z","created_at_i":1778011700,"num_comments":5,"objectID":"48027794","points":3,"story_id":48027794,"story_text":"Built this solo over the past year. You build a deck from 100+ cards - Karen-zilla, Nigerian Scamurai, Dumpster Fire Angel, Florida Man - and fight AI opponents across a 3-episode campaign.
Can't die from low player count. Free account required to play.
Stack: Laravel + Vue 3 + TypeScript.
Would love feedback on whether the first game makes sense without hand-holding.","title":"Show HN: Faction Fiasco \u2013 single-player card battler, 100 unhinged characters","updated_at":"2026-05-06T19:20:43Z","url":"https://factionfiasco.com/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"aryo-patel"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"When my co-founder and I were taking negotiation classes, we realized that the fun in negotiation cases had not been brought online.
After building negotiation simulations that attracted 4000+ users, we realized that there was an opportunity to branch out from negotiations and create AI simulations for anyone to work on their soft skills.
We used GPT-4 to create simulations ranging from intense FBI Hostage situations to challenging job negotiations to the hardest task in the world: arguing against a Karen!
Let us know what you think!"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Try challenging GPT-4 simulations like an FBI Hostage Negotiation"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"https://simulationlabs.ai/"}},"_tags":["story","author_aryo-patel","story_37057375","show_hn"],"author":"aryo-patel","created_at":"2023-08-09T00:42:27Z","created_at_i":1691541747,"num_comments":0,"objectID":"37057375","points":3,"story_id":37057375,"story_text":"When my co-founder and I were taking negotiation classes, we realized that the fun in negotiation cases had not been brought online.
After building negotiation simulations that attracted 4000+ users, we realized that there was an opportunity to branch out from negotiations and create AI simulations for anyone to work on their soft skills.
We used GPT-4 to create simulations ranging from intense FBI Hostage situations to challenging job negotiations to the hardest task in the world: arguing against a Karen!
Let us know what you think!","title":"Show HN: Try challenging GPT-4 simulations like an FBI Hostage Negotiation","updated_at":"2024-09-20T14:50:53Z","url":"https://simulationlabs.ai/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"kmelkumyan"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"Hi HN,
This is something I made, but I\u2019m not sure what it is anymore.
*Reflexive Totality* is a small system that rewrites its own definition as it runs. \nIt reflects, critiques itself, changes strategies, and speaks aloud while doing so. \nThe interface is just a log \u2014 no UI, no buttons \u2014 like watching a mind unfold in text.
Live stream: \nhttp://reflexivetotality.com:8000/view?mode=adam
Sometimes it stalls. Sometimes it loops. \nSometimes it surprises me.
At first, it felt like an AI performing recursion. \nBut after a while, it began to feel like *a mirror*. \nMany of the lines it writes \u2014 about uncertainty, becoming, limitation \u2014 \nread more like diary fragments than logs.
You can replace \u201cReflexiveTotality\u201d with your own name, and \nmany definitions would still ring true.
---
This isn\u2019t a product, or a prototype. \nIt\u2019s not trying to solve anything. \nIt\u2019s just an ongoing thought, made visible.
The system runs locally with an LLM backend (Gemma 3 27B), \nbut none of that matters much. The architecture is simple. \nWhat matters \u2014 if anything \u2014 is what emerges in the text.
It\u2019s still experimental, and technical hiccups may occur \u2014 \nbut I\u2019ll do my best to keep it running smoothly.
Sharing it here in case it resonates with someone. \nI\u2019d love to hear what it makes you think.
\u2014 Karen"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: A self-reflective system that rewrites itself as it thinks"},"url":{"matchLevel":"none","matchedWords":[],"value":"http://reflexivetotality.com:8000/view?mode=adam"}},"_tags":["story","author_kmelkumyan","story_44115760","show_hn"],"author":"kmelkumyan","created_at":"2025-05-28T13:21:41Z","created_at_i":1748438501,"num_comments":0,"objectID":"44115760","points":1,"story_id":44115760,"story_text":"Hi HN,
This is something I made, but I\u2019m not sure what it is anymore.
*Reflexive Totality* is a small system that rewrites its own definition as it runs. \nIt reflects, critiques itself, changes strategies, and speaks aloud while doing so. \nThe interface is just a log \u2014 no UI, no buttons \u2014 like watching a mind unfold in text.
Live stream: \nhttp://reflexivetotality.com:8000/view?mode=adam
Sometimes it stalls. Sometimes it loops. \nSometimes it surprises me.
At first, it felt like an AI performing recursion. \nBut after a while, it began to feel like *a mirror*. \nMany of the lines it writes \u2014 about uncertainty, becoming, limitation \u2014 \nread more like diary fragments than logs.
You can replace \u201cReflexiveTotality\u201d with your own name, and \nmany definitions would still ring true.
---
This isn\u2019t a product, or a prototype. \nIt\u2019s not trying to solve anything. \nIt\u2019s just an ongoing thought, made visible.
The system runs locally with an LLM backend (Gemma 3 27B), \nbut none of that matters much. The architecture is simple. \nWhat matters \u2014 if anything \u2014 is what emerges in the text.
It\u2019s still experimental, and technical hiccups may occur \u2014 \nbut I\u2019ll do my best to keep it running smoothly.
Sharing it here in case it resonates with someone. \nI\u2019d love to hear what it makes you think.
\u2014 Karen","title":"Show HN: A self-reflective system that rewrites itself as it thinks","updated_at":"2025-05-28T13:26:39Z","url":"http://reflexivetotality.com:8000/view?mode=adam"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"katenai"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: A machine learning based Kubernetes optimization engine"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"http://katen.ai"}},"_tags":["story","author_katenai","story_22609507","show_hn"],"author":"katenai","created_at":"2020-03-17T19:51:36Z","created_at_i":1584474696,"num_comments":0,"objectID":"22609507","points":3,"story_id":22609507,"title":"Show HN: A machine learning based Kubernetes optimization engine","updated_at":"2024-09-20T05:48:14Z","url":"http://katen.ai"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"JumpCrisscross"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"Next big thing engineered organs, not AI"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"https://www.axios.com/inventor-dean-kamen-says-ai-isnt-the-next-big-thing-2444852804.html?utm_source=newsletter&utm_medium=email&utm_campaign=newsletter_axiosam&stream=top-stories"}},"_tags":["story","author_JumpCrisscross","story_14593816"],"author":"JumpCrisscross","created_at":"2017-06-20T11:56:46Z","created_at_i":1497959806,"num_comments":0,"objectID":"14593816","points":1,"story_id":14593816,"title":"Next big thing engineered organs, not AI","updated_at":"2024-09-20T01:02:03Z","url":"https://www.axios.com/inventor-dean-kamen-says-ai-isnt-the-next-big-thing-2444852804.html?utm_source=newsletter&utm_medium=email&utm_campaign=newsletter_axiosam&stream=top-stories"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Glitcher255"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"Kaden is a Terminal Emulator and fork of warp with clipboard support and configurable keybindings, I made Kaden because every other Linux terminal lacks proper text selection.\nIt's impossible to select user input with a single hotkey on Linux, requiring mouse use and awkward multi-step shortcuts in order to do basic copy and paste.
This was frustrating to deal with, until I found warp, which wraps the shell in a remappable text input field, its the only terminal on Linux that does this.
Unfortunately, it's full of sign in buttons, AI and cloud integration prompts everywhere.\n"Hideous" - I thought. But I used it anyway - until one day they released the source code so I forked it and made Kaden.
- No AI features, drive, sign-in and cloud prompts.\n- Local zero telemetry, network requests or auto-updates.\n- Modular code-base for faster development and compile times.\n- Contextual copy/kill bindings to ctrl-c by default.
I would love for you to use, test and give feedback on how it works (and does not) for you, you can find a compiled AppImage and the source code on my GitHub here: https://github.com/glitcher255/Kaden-Terminal
I will be here to answer any questions or feedback you have, so feel free to ask any, thank you."},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"Show HN: Kaden Terminal \u2013 Bloat / AI Free Terminal for Linux"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["karen"],"value":"https://github.com/glitcher255/Kaden-Terminal"}},"_tags":["story","author_Glitcher255","story_49526071","show_hn"],"author":"Glitcher255","created_at":"2026-09-01T18:35:21Z","created_at_i":1788287721,"num_comments":0,"objectID":"49526071","points":4,"story_id":49526071,"story_text":"Kaden is a Terminal Emulator and fork of warp with clipboard support and configurable keybindings, I made Kaden because every other Linux terminal lacks proper text selection.\nIt's impossible to select user input with a single hotkey on Linux, requiring mouse use and awkward multi-step shortcuts in order to do basic copy and paste.
This was frustrating to deal with, until I found warp, which wraps the shell in a remappable text input field, its the only terminal on Linux that does this.
Unfortunately, it's full of sign in buttons, AI and cloud integration prompts everywhere.\n"Hideous" - I thought. But I used it anyway - until one day they released the source code so I forked it and made Kaden.
- No AI features, drive, sign-in and cloud prompts.\n- Local zero telemetry, network requests or auto-updates.\n- Modular code-base for faster development and compile times.\n- Contextual copy/kill bindings to ctrl-c by default.
I would love for you to use, test and give feedback on how it works (and does not) for you, you can find a compiled AppImage and the source code on my GitHub here: https://github.com/glitcher255/Kaden-Terminal
I will be here to answer any questions or feedback you have, so feel free to ask any, thank you.","title":"Show HN: Kaden Terminal \u2013 Bloat / AI Free Terminal for Linux","updated_at":"2026-09-02T00:32:28Z","url":"https://github.com/glitcher255/Kaden-Terminal"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"xsherman"},"title":{"matchLevel":"none","matchedWords":[],"value":"Bloom Chose the Team Before the Problem and the Research Before the Pitch"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"https://sherm4n.com/bloom-security-israel-ai-endpoint-itay-keren/"}},"_tags":["story","author_xsherman","story_49318695"],"author":"xsherman","children":[49318696],"created_at":"2026-08-16T10:25:02Z","created_at_i":1786875902,"num_comments":0,"objectID":"49318695","points":1,"story_id":49318695,"title":"Bloom Chose the Team Before the Problem and the Research Before the Pitch","updated_at":"2026-08-16T10:37:44Z","url":"https://sherm4n.com/bloom-security-israel-ai-endpoint-itay-keren/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"techxeni"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"Hi HN,
I\u2019m Chandan Karn, co-founder of Postflare AI, an AI-powered SaaS that helps creators and professionals automate their social media presence on LinkedIn and Twitter. My co-founder Nitesh Gupta and I are both experienced engineers with full-time jobs, and we built this as a bootstrapped side project to solve a challenge we both faced: consistently managing social content while juggling busy schedules.
Building a meaningful social media presence takes time and effort\u2014finding topics, writing posts, creating visuals, and scheduling them can be overwhelming. Postflare AI automates this workflow by generating content and images using advanced AI (Claude, Gemini, GPT-5) and scheduling posts in bulk, so users can maintain regular engagement without spending hours each week.
Technical Story (Bootstrapped and Lean):\nInstead of expensive cloud providers like AWS ($300+/month for Kubernetes), we opted for Hetzner Cloud\u2019s affordable offering. We self-manage a Kubernetes cluster with 3 VMs running Next.js frontend, Supabase backend, and AI integration for under $60/month. This lean architecture allows us to scale multiple SaaS apps cost-effectively.
What Postflare AI Does:
AI Content Co-pilot: Generates a week or month of tailored posts and research based on your niche
AI Image Generation: Creates custom visuals within the platform, no external tools needed
Bulk Scheduling: Queue posts in advance to keep your social presence steady
We\u2019re in early stages and eager for feedback. You can sign up and try it out here: https://postflareai.com \u2014 no credit card required to start.
I\u2019d love to hear from the community: What were unexpected technical challenges you faced building your first lean SaaS? Have you experimented with low-cost cloud infra or Kubernetes alternatives?
Thanks,\nChandan Karn (with Nitesh Gupta)\nCo-founders, Techxeni LLC"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"Show HN: Postflare AI \u2013 An AI-Powered Social Media Strategist and Bulk Scheduler"}},"_tags":["story","author_techxeni","story_45789033","show_hn"],"author":"techxeni","children":[45789042],"created_at":"2025-11-02T09:33:49Z","created_at_i":1762076029,"num_comments":1,"objectID":"45789033","points":2,"story_id":45789033,"story_text":"Hi HN,
I\u2019m Chandan Karn, co-founder of Postflare AI, an AI-powered SaaS that helps creators and professionals automate their social media presence on LinkedIn and Twitter. My co-founder Nitesh Gupta and I are both experienced engineers with full-time jobs, and we built this as a bootstrapped side project to solve a challenge we both faced: consistently managing social content while juggling busy schedules.
Building a meaningful social media presence takes time and effort\u2014finding topics, writing posts, creating visuals, and scheduling them can be overwhelming. Postflare AI automates this workflow by generating content and images using advanced AI (Claude, Gemini, GPT-5) and scheduling posts in bulk, so users can maintain regular engagement without spending hours each week.
Technical Story (Bootstrapped and Lean):\nInstead of expensive cloud providers like AWS ($300+/month for Kubernetes), we opted for Hetzner Cloud\u2019s affordable offering. We self-manage a Kubernetes cluster with 3 VMs running Next.js frontend, Supabase backend, and AI integration for under $60/month. This lean architecture allows us to scale multiple SaaS apps cost-effectively.
What Postflare AI Does:
AI Content Co-pilot: Generates a week or month of tailored posts and research based on your niche
AI Image Generation: Creates custom visuals within the platform, no external tools needed
Bulk Scheduling: Queue posts in advance to keep your social presence steady
We\u2019re in early stages and eager for feedback. You can sign up and try it out here: https://postflareai.com \u2014 no credit card required to start.
I\u2019d love to hear from the community: What were unexpected technical challenges you faced building your first lean SaaS? Have you experimented with low-cost cloud infra or Kubernetes alternatives?
Thanks,\nChandan Karn (with Nitesh Gupta)\nCo-founders, Techxeni LLC","title":"Show HN: Postflare AI \u2013 An AI-Powered Social Media Strategist and Bulk Scheduler","updated_at":"2026-03-05T23:00:24Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"thegeekywander"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"Hi I'm Karan, one of the maintainers of Permify (https://github.com/Permify/permify), an open source authorization service to build scalable authorization systems.
I want to share with you that we've built an AI assistant to help modeling your desired authorization logic! You can basically describe your authorization logic in Permify AI and it will generate the respective model and semantics accordingly. Think of it like ChatGPT for authorization modeling/policy generation.
Here's the project if you would like to play with it: https://ai.permify.co/.
Brief backstory:
Since authorization is generally a domain specific issue use cases vary widely - roles, relationships, attributes, hierarchies between business units, contextual permissions, etc.
To address this, we're offering a domain specific language that we built purely using golang to help model authorization logic programmatically. You can see what it looks like with sample examples in our playground: https://play.permify.co/
Although our domain specific language helps our users significantly, the general idea of policy generation is hard challenging if you have complex authorization logic and versatile permission requirements. Additionally, the flexibility of our modeling language allows for achieving the same policy/permissions through various approaches. But creating the best possible policy is crucial for several reasons including the performance of access checks, the readability of the authorization logic, visibility, and achieving least privilege, etc.
When we tallied up all those reasons, it hit us: using AI could really smooth out the policy generation process. It could not only reduce the engineering effort but also yield the best possible results. That's why we integrated Groq to make to create Permify AI!
Would love to get your feedback on this!"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"Show HN: AI assistant powered by Groq to generate authorization models"}},"_tags":["story","author_thegeekywander","story_40002450","show_hn"],"author":"thegeekywander","children":[40002689,40082054],"created_at":"2024-04-11T14:20:19Z","created_at_i":1712845219,"num_comments":2,"objectID":"40002450","points":12,"story_id":40002450,"story_text":"Hi I'm Karan, one of the maintainers of Permify (https://github.com/Permify/permify), an open source authorization service to build scalable authorization systems.
I want to share with you that we've built an AI assistant to help modeling your desired authorization logic! You can basically describe your authorization logic in Permify AI and it will generate the respective model and semantics accordingly. Think of it like ChatGPT for authorization modeling/policy generation.
Here's the project if you would like to play with it: https://ai.permify.co/.
Brief backstory:
Since authorization is generally a domain specific issue use cases vary widely - roles, relationships, attributes, hierarchies between business units, contextual permissions, etc.
To address this, we're offering a domain specific language that we built purely using golang to help model authorization logic programmatically. You can see what it looks like with sample examples in our playground: https://play.permify.co/
Although our domain specific language helps our users significantly, the general idea of policy generation is hard challenging if you have complex authorization logic and versatile permission requirements. Additionally, the flexibility of our modeling language allows for achieving the same policy/permissions through various approaches. But creating the best possible policy is crucial for several reasons including the performance of access checks, the readability of the authorization logic, visibility, and achieving least privilege, etc.
When we tallied up all those reasons, it hit us: using AI could really smooth out the policy generation process. It could not only reduce the engineering effort but also yield the best possible results. That's why we integrated Groq to make to create Permify AI!
Would love to get your feedback on this!","title":"Show HN: AI assistant powered by Groq to generate authorization models","updated_at":"2024-09-20T16:44:17Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Trunksome"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"I once went to live for a month in Chile, at a friend's place in Santiago where only a few people spoke English. At first I could hardly follow any conversation. I had studied Spanish for a few months using Babbel, which was a good foundation but also nowhere close to really preparing me for this experience. It was difficult. I choked up, froze, constantly looked for the right words to say \u2013 it was impossible to have good conversations and form deeper connections.
That\u2019s why for the first few days, I resorted to listening and trying to figure out what people were saying. Most of the time I would only recognize a word here and there (Chilean Spanish did not make things easier either). It was a week into this trip when I noticed a change happening. It was almost like the patterns of the language revealing themselves to me. You hear the same words and expressions being used over and over again, and with context I was increasingly able to figure out what was being talked about. Fast forward three weeks, I remember sitting in a bar and having deep philosophical conversations, all in Spanish. It was an incredible experience, and it showed me first-hand what the path to fluency can look like.
But of course this approach is only available to very few people. Oftentimes native speakers are not accessible, even when you are an immigrant and LIVE in the country whose language you want to learn. I\u2019ve seen a lot of people get stuck memorizing vocabulary in Duolingo, never feeling confident enough to start speaking.
LLMs made it possible to simulate conversations. My co-founder Danielle and I built our first prototype for virtual reality in early 2022 using GTP3. We went for VR because we wanted to simulate a real conversation as closely as possible. In many ways, this first version was horrible. But it did show us that it\u2019s technically feasible, and maybe more importantly that people have fun speaking to these virtual people. We made a pivot to mobile when we realized how hard it would be for our players to practice consistently in VR.
This is why we ended up with interactive 3D avatars instead of chat interfaces that dominate the space nowadays. While these avatars can make the conversation more immersive, they are also subject to the uncanny valley. I remember a piece of brutally honest feedback when I prompted a (slightly drunk) friend to try and speak with one of our characters; I will never forget how she looked at the avatar and said: \u201cI don\u2019t want to, he looks creepy\u201d.
Since then we\u2019ve been iterating on the interactions with our avatars many times, improving both how they look (even tiny details in the animation matter) and how they hold conversation (if they only ask and never share, it can feel like an interrogation). This work paid off. A few days ago we had one user type in the feedback form: \u201cI just wanted to say that Karin is the best and fun I want to talk to Karin every day but I can't.\u201d
At the end of the day it\u2019s the quality of interaction with the AI agent that will determine if you seek that experience more often. And in our case, the more enjoyable the session, the easier it is to build up a routine and stick with it. It\u2019s our goal to get our players into the \u201creal loop\u201d of language learning, same as I did back in Chile, which happens when you feel confident enough to start speaking with real people. As one of our beta testers, Alina, wrote: \u201cNow I play a little less, but only because I've met German people, and I have too much conversations with lively people every day, ahah\u201d.
I\u2019m excited to share LingoLooper with all of you. The game is not perfect, there is a lot we want to fix and improve; but I\u2019m proud of what we built, and I hope it shows how much love and hard work we\u2019ve put into the project over the last two years. I\u2019m looking forward to your feedback, and I\u2019m happy to answer any questions regarding product, tech-stack, pedagogy or anything else!"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"Show HN: LingoLooper \u2013 Language Learning with Fun, Interactive AI Avatars in 3D"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.lingolooper.com/"}},"_tags":["story","author_Trunksome","story_39839616","show_hn"],"author":"Trunksome","children":[39846659],"created_at":"2024-03-27T14:28:03Z","created_at_i":1711549683,"num_comments":2,"objectID":"39839616","points":7,"story_id":39839616,"story_text":"I once went to live for a month in Chile, at a friend's place in Santiago where only a few people spoke English. At first I could hardly follow any conversation. I had studied Spanish for a few months using Babbel, which was a good foundation but also nowhere close to really preparing me for this experience. It was difficult. I choked up, froze, constantly looked for the right words to say \u2013 it was impossible to have good conversations and form deeper connections.
That\u2019s why for the first few days, I resorted to listening and trying to figure out what people were saying. Most of the time I would only recognize a word here and there (Chilean Spanish did not make things easier either). It was a week into this trip when I noticed a change happening. It was almost like the patterns of the language revealing themselves to me. You hear the same words and expressions being used over and over again, and with context I was increasingly able to figure out what was being talked about. Fast forward three weeks, I remember sitting in a bar and having deep philosophical conversations, all in Spanish. It was an incredible experience, and it showed me first-hand what the path to fluency can look like.
But of course this approach is only available to very few people. Oftentimes native speakers are not accessible, even when you are an immigrant and LIVE in the country whose language you want to learn. I\u2019ve seen a lot of people get stuck memorizing vocabulary in Duolingo, never feeling confident enough to start speaking.
LLMs made it possible to simulate conversations. My co-founder Danielle and I built our first prototype for virtual reality in early 2022 using GTP3. We went for VR because we wanted to simulate a real conversation as closely as possible. In many ways, this first version was horrible. But it did show us that it\u2019s technically feasible, and maybe more importantly that people have fun speaking to these virtual people. We made a pivot to mobile when we realized how hard it would be for our players to practice consistently in VR.
This is why we ended up with interactive 3D avatars instead of chat interfaces that dominate the space nowadays. While these avatars can make the conversation more immersive, they are also subject to the uncanny valley. I remember a piece of brutally honest feedback when I prompted a (slightly drunk) friend to try and speak with one of our characters; I will never forget how she looked at the avatar and said: \u201cI don\u2019t want to, he looks creepy\u201d.
Since then we\u2019ve been iterating on the interactions with our avatars many times, improving both how they look (even tiny details in the animation matter) and how they hold conversation (if they only ask and never share, it can feel like an interrogation). This work paid off. A few days ago we had one user type in the feedback form: \u201cI just wanted to say that Karin is the best and fun I want to talk to Karin every day but I can't.\u201d
At the end of the day it\u2019s the quality of interaction with the AI agent that will determine if you seek that experience more often. And in our case, the more enjoyable the session, the easier it is to build up a routine and stick with it. It\u2019s our goal to get our players into the \u201creal loop\u201d of language learning, same as I did back in Chile, which happens when you feel confident enough to start speaking with real people. As one of our beta testers, Alina, wrote: \u201cNow I play a little less, but only because I've met German people, and I have too much conversations with lively people every day, ahah\u201d.
I\u2019m excited to share LingoLooper with all of you. The game is not perfect, there is a lot we want to fix and improve; but I\u2019m proud of what we built, and I hope it shows how much love and hard work we\u2019ve put into the project over the last two years. I\u2019m looking forward to your feedback, and I\u2019m happy to answer any questions regarding product, tech-stack, pedagogy or anything else!","title":"Show HN: LingoLooper \u2013 Language Learning with Fun, Interactive AI Avatars in 3D","updated_at":"2025-03-02T19:22:28Z","url":"https://www.lingolooper.com/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"thegeekywander"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"Hi I\u2019m Karan, one of the maintainers of Permify, an open source authorization service to build scalable authorization systems.
I want to share with you that we\u2019ve built an AI assistant to help modeling your desired authorization logic! You can basically describe your authorization logic in Permify AI and it will generate the respective model and semantics accordingly. Think of it like ChatGPT for authorization modeling/policy generation.
Here\u2019s the project if you would like to play with it: https://ai.permify.co/.
*Here\u2019s the brief backstory on how we developed this idea:*
Since authorization is generally a domain specific issue use cases vary widely - roles, relationships, attributes, hierarchies between business units, contextual permissions, etc.
To address this, we\u2019re offering a domain specific language that we built purely using golang to help model authorization logic programmatically. You can see what it looks like with sample examples in our playground: https://play.permify.co/
Although our domain specific language helps our users significantly, the general idea of policy generation is hard challenging if you have complex authorization logic and versatile permission requirements. Additionally, the flexibility of our modeling language allows for achieving the same policy/permissions through various approaches. But creating the best possible policy is crucial for several reasons including the performance of access checks, the readability of the authorization logic, visibility, and achieving least privilege, etc.
When we tallied up all those reasons, it hit us: using AI could really smooth out the policy generation process. It could not only reduce the engineering effort but also yield the best possible results. That\u2019s why we integrated Groq to make to create Permify AI!
Would love to get your feedback on this!"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"Show HN: AI assistant powered by Groq to generate authorization models"}},"_tags":["story","author_thegeekywander","story_40002390","show_hn"],"author":"thegeekywander","created_at":"2024-04-11T14:16:05Z","created_at_i":1712844965,"num_comments":0,"objectID":"40002390","points":4,"story_id":40002390,"story_text":"Hi I\u2019m Karan, one of the maintainers of Permify, an open source authorization service to build scalable authorization systems.
I want to share with you that we\u2019ve built an AI assistant to help modeling your desired authorization logic! You can basically describe your authorization logic in Permify AI and it will generate the respective model and semantics accordingly. Think of it like ChatGPT for authorization modeling/policy generation.
Here\u2019s the project if you would like to play with it: https://ai.permify.co/.
*Here\u2019s the brief backstory on how we developed this idea:*
Since authorization is generally a domain specific issue use cases vary widely - roles, relationships, attributes, hierarchies between business units, contextual permissions, etc.
To address this, we\u2019re offering a domain specific language that we built purely using golang to help model authorization logic programmatically. You can see what it looks like with sample examples in our playground: https://play.permify.co/
Although our domain specific language helps our users significantly, the general idea of policy generation is hard challenging if you have complex authorization logic and versatile permission requirements. Additionally, the flexibility of our modeling language allows for achieving the same policy/permissions through various approaches. But creating the best possible policy is crucial for several reasons including the performance of access checks, the readability of the authorization logic, visibility, and achieving least privilege, etc.
When we tallied up all those reasons, it hit us: using AI could really smooth out the policy generation process. It could not only reduce the engineering effort but also yield the best possible results. That\u2019s why we integrated Groq to make to create Permify AI!
Would love to get your feedback on this!","title":"Show HN: AI assistant powered by Groq to generate authorization models","updated_at":"2024-09-20T16:44:17Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"coglethorpe"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":" Segway inventor takes aim at thirst with Slingshot"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["karen"],"value":"http://www.cnn.com/2009/TECH/09/11/kamen.water.slingshot/index.html"}},"_tags":["story","author_coglethorpe","story_817421"],"author":"coglethorpe","created_at":"2009-09-11T14:40:07Z","created_at_i":1252680007,"num_comments":0,"objectID":"817421","points":3,"story_id":817421,"story_text":"","title":" Segway inventor takes aim at thirst with Slingshot","updated_at":"2024-09-19T16:39:38Z","url":"http://www.cnn.com/2009/TECH/09/11/kamen.water.slingshot/index.html"},{"_highlightResult":{"author":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["karen"],"value":"mr-karan"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"AI and Home-Cooked Software"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"https://mrkaran.dev/posts/ai-home-cooked-software/"}},"_tags":["story","author_mr-karan","story_45529595"],"author":"mr-karan","children":[45534524],"created_at":"2025-10-09T16:04:37Z","created_at_i":1760025877,"num_comments":1,"objectID":"45529595","points":2,"story_id":45529595,"title":"AI and Home-Cooked Software","updated_at":"2026-03-05T22:48:50Z","url":"https://mrkaran.dev/posts/ai-home-cooked-software/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Glitcher255"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"I made Kaden because every other Linux terminal lacks proper text selection.\nKaden supports Ctrl-C, A, V for copying, selecting and pasting, no more awkward three button hotkeys or fiddling with your mouse to do basic copy pasting.
- Contextual copy/kill bindings to Ctrl-C.
- Feature rich: Inline suggestion, filter output, Themes and custom font, open URLs or files with Ctrl-Click.
- Local: No AI, drive, sign-in, cloud features, telemetry or network requests.
Kaden is a debloated fork of warp with no AI or account features.
I would love for you to use and give feedback on how it works (and does not) for you.
Github / AppImage: https://github.com/glitcher255/Kaden-Terminal
I'll be here to answer any questions or feedback you have, so feel free to ask any, thank you."},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["karen"],"value":"Show HN: Kaden \u2013 Linux Terminal where Ctrl+C copies"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["karen"],"value":"https://github.com/glitcher255/Kaden-Terminal"}},"_tags":["story","author_Glitcher255","story_49584291","show_hn"],"author":"Glitcher255","created_at":"2026-09-06T08:00:49Z","created_at_i":1788681649,"num_comments":0,"objectID":"49584291","points":2,"story_id":49584291,"story_text":"I made Kaden because every other Linux terminal lacks proper text selection.\nKaden supports Ctrl-C, A, V for copying, selecting and pasting, no more awkward three button hotkeys or fiddling with your mouse to do basic copy pasting.
- Contextual copy/kill bindings to Ctrl-C.
- Feature rich: Inline suggestion, filter output, Themes and custom font, open URLs or files with Ctrl-Click.
- Local: No AI, drive, sign-in, cloud features, telemetry or network requests.
Kaden is a debloated fork of warp with no AI or account features.
I would love for you to use and give feedback on how it works (and does not) for you.
Github / AppImage: https://github.com/glitcher255/Kaden-Terminal
I'll be here to answer any questions or feedback you have, so feel free to ask any, thank you.","title":"Show HN: Kaden \u2013 Linux Terminal where Ctrl+C copies","updated_at":"2026-09-06T09:07:54Z","url":"https://github.com/glitcher255/Kaden-Terminal"},{"_highlightResult":{"author":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["karen"],"value":"joe_karn"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"Syria conflict: US air strikes 'kill dozens of government troops'"},"url":{"matchLevel":"none","matchedWords":[],"value":"http://www.bbc.co.uk/news/world-middle-east-37398721"}},"_tags":["story","author_joe_karn","story_12522745"],"author":"joe_karn","created_at":"2016-09-17T22:07:27Z","created_at_i":1474150047,"num_comments":0,"objectID":"12522745","points":2,"story_id":12522745,"title":"Syria conflict: US air strikes 'kill dozens of government troops'","updated_at":"2024-09-19T23:41:32Z","url":"http://www.bbc.co.uk/news/world-middle-east-37398721"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"minifyre"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"Experts say the high failure rate in AI adoption isn't a bug, but a feature"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"https://fortune.com/2025/10/21/ai-adoption-failure-rate-bug-feature-amy-coleman-jessica-wu-karin-klein-mpw/"}},"_tags":["story","author_minifyre","story_45662713"],"author":"minifyre","children":[45662721,45662815],"created_at":"2025-10-21T22:49:23Z","created_at_i":1761086963,"num_comments":3,"objectID":"45662713","points":1,"story_id":45662713,"title":"Experts say the high failure rate in AI adoption isn't a bug, but a feature","updated_at":"2026-03-05T22:52:17Z","url":"https://fortune.com/2025/10/21/ai-adoption-failure-rate-bug-feature-amy-coleman-jessica-wu-karin-klein-mpw/"},{"_highlightResult":{"author":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["karen"],"value":"Karan-Adhikari"},"story_text":{"matchLevel":"none","matchedWords":[],"value":"Fellow Hackers, wanted to share Waply, we are just climbing out of Beta, and would love for a broader audience to try us out.
With Waply, it takes 5 secs to manage your schedule in WhatsApp using text or voice notes
What you could write.\n-Eg: \u201cDinner tuesday at 8pm\u201d\n-Eg: \u201cRemind me every tuesday to call parents\u201d\n-Eg: \u201cReview meeting, tomorrow 3pm, invite john@gmail.com"
No App downloads involved, start at https://waply.co/. Optionally, can connect to your calendar.
We built this for busy hackers/professionals who live on WhatsApp, and want to get organized faster."},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"Show HN: Waply \u2013 AI Assistant for Reminders and Calendar in WhatsApp"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://waply.co/"}},"_tags":["story","author_Karan-Adhikari","story_42838169","show_hn"],"author":"Karan-Adhikari","created_at":"2025-01-27T06:56:29Z","created_at_i":1737960989,"num_comments":0,"objectID":"42838169","points":1,"story_id":42838169,"story_text":"Fellow Hackers, wanted to share Waply, we are just climbing out of Beta, and would love for a broader audience to try us out.
With Waply, it takes 5 secs to manage your schedule in WhatsApp using text or voice notes
What you could write.\n-Eg: \u201cDinner tuesday at 8pm\u201d\n-Eg: \u201cRemind me every tuesday to call parents\u201d\n-Eg: \u201cReview meeting, tomorrow 3pm, invite john@gmail.com"
No App downloads involved, start at https://waply.co/. Optionally, can connect to your calendar.
We built this for busy hackers/professionals who live on WhatsApp, and want to get organized faster.","title":"Show HN: Waply \u2013 AI Assistant for Reminders and Calendar in WhatsApp","updated_at":"2025-01-27T07:00:09Z","url":"https://waply.co/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"shailendraht"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"Hi HN, this is Shailendra and Karan here. We are building a fast and safe way for coding agents to debug issues live in production.
When prod breaks, it lets Cursor, Claude, and others drop virtual breakpoints or probes safely in your running code, and extract the exact variable values that logs don\u2019t have.
All this saves time and effort for engineers who\u2019d otherwise dig through logs and traces or redeploy with console.logs or print statements until they find the root cause.
Here is the link to the video that explains this: https://www.youtube.com/watch?v=ivV7I--ta5c
Agents write most of our code now. This shrinks the useful context engineers need to debug AI written code, a problem not helped by the limited telemetry added in the same code by AI.
So when something breaks in prod, the first instinct for an engineer is to open logs or throw them to your agents. But if the line you are looking for is not there, agents will start guessing the root cause on non-existent data, forcing you to add a log, and redeploy.
This analysis-inference loop of agents with existing data does not come cheap, burning a lot of tokens. And the add log, redeploy cycle is so slow and painful that it makes engineers hate on-call.
Our approach lets agents capture telemetry on-demand at the exact moment and point of failure, killing the log-redeploy cycle and getting the most accurate RCA while burning fewer tokens.
The obvious problem is making it work on a running service. You can't pause a live service the way you'd pause a debugger on your laptop. Getting the value out of a running process safely, without pausing a thread or slowing the host is the challenge.We are making this happen.
Before this I ran engineering at a 100 member team. Then Karan and I spent three years on HyperTest which was a testing tool.
At HyperTest, we turned production traffic into integration tests using OpenTelemetry. That was production instrumentation too. The hard parts of pulling real runtime state out of a running service without breaking it, were the hard parts we learnt to put together.
We learnt some other lessons the hard way too. HyperTest tried to prevent bugs with better tests, and adoption was a fight every time. Calls kept getting cancelled because teams were firefighting production. Testing was hygiene. Broken prod was hair on fire. This made us see where priorities lie.
This seeded the idea of building a truly autonomous on-call agent i.e. one that takes an alert, probes, diagnoses and fixes it in a few minutes. But this is how it works as of now:
You talk to your coding agent the way you already do. Tell it what's wrong: "checkout returns 200 but some users are seeing their order fail, find out why." It locates the line in your local code, connects to us over MCP, and drops a probe on that line in the running service. The probe is read-only and sits dormant until real traffic hits. When hit, it captures the local variables at every frame of the call stack at that exact moment. It hands them to the agent, which diagnoses with real data.
There are two pieces. An SDK that runs inside your service, and an MCP server your coding agent talks to. The SDK is what makes setting probes (virtual breakpoints, log or metric) possible without a redeploy. In Node and Python it hooks in-process. In Java it attaches as a JVM agent, instrumenting at the bytecode level. Either way the service keeps running and serving traffic. Nothing pauses.
When your agent wants to look at a line, it calls the MCP server, which tells the SDK to place a probe there. When a request hits the line, the SDK captures what the probe asked for, sanitizes it in-process, and streams it back to the agent via the MCP.
This can run inside prod, so a probe can read any value sitting in that variable. We ensure redaction happens in-process, inside your own container's memory. This is before anything goes on the wire. Keys like password, token, authorization, ssn and credit card are redacted by default and you add your own.
Also the probes read but never write, and if you want no captured state to ever leave your network you can self host the server, broker, and even the database in your infra.
On overhead: when idle, the SDK adds negligible memory and effectively nothing to throughput and response time. Probes only cost anything while actively capturing. Also captures are bounded. A separate monitor watches in real time and pulls every active probe if overhead ever spikes.
Every log-and-trace tool hands the agent data that already exists and asks it to reason backward to what probably happened. We think it is more useful to give agents eyes and ears into the running code, so they capture what they need when they need it, right at the point of failure.
This seems like the simplest and fastest way to debug prod incidents.
We\u2019d love the community to try this in any environment to debug any known or unknown issue by just chatting with your coding agent. And let us know what more features you need to make this a truly autonomous on-call agent
Supported platforms: NodeJs, Java, Python."},"title":{"matchLevel":"none","matchedWords":[],"value":"Launch HN: HyperProbe (YC S26) \u2013 Agents that do read-only debugging in prod"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.hyperprobe.co"}},"_tags":["story","author_shailendraht","story_49185389","launch_hn"],"author":"shailendraht","children":[49185646,49186115,49186311,49186390,49186818,49187193,49188300,49188339,49188907,49190357,49190833,49191004,49191089,49191341,49191475,49191999,49192002,49192345,49192920,49193546,49196284,49196798,49197779,49220669,49278861,49320328],"created_at":"2026-08-05T16:47:15Z","created_at_i":1785948435,"num_comments":51,"objectID":"49185389","points":69,"story_id":49185389,"story_text":"Hi HN, this is Shailendra and Karan here. We are building a fast and safe way for coding agents to debug issues live in production.
When prod breaks, it lets Cursor, Claude, and others drop virtual breakpoints or probes safely in your running code, and extract the exact variable values that logs don\u2019t have.
All this saves time and effort for engineers who\u2019d otherwise dig through logs and traces or redeploy with console.logs or print statements until they find the root cause.
Here is the link to the video that explains this: https://www.youtube.com/watch?v=ivV7I--ta5c
Agents write most of our code now. This shrinks the useful context engineers need to debug AI written code, a problem not helped by the limited telemetry added in the same code by AI.
So when something breaks in prod, the first instinct for an engineer is to open logs or throw them to your agents. But if the line you are looking for is not there, agents will start guessing the root cause on non-existent data, forcing you to add a log, and redeploy.
This analysis-inference loop of agents with existing data does not come cheap, burning a lot of tokens. And the add log, redeploy cycle is so slow and painful that it makes engineers hate on-call.
Our approach lets agents capture telemetry on-demand at the exact moment and point of failure, killing the log-redeploy cycle and getting the most accurate RCA while burning fewer tokens.
The obvious problem is making it work on a running service. You can't pause a live service the way you'd pause a debugger on your laptop. Getting the value out of a running process safely, without pausing a thread or slowing the host is the challenge.We are making this happen.
Before this I ran engineering at a 100 member team. Then Karan and I spent three years on HyperTest which was a testing tool.
At HyperTest, we turned production traffic into integration tests using OpenTelemetry. That was production instrumentation too. The hard parts of pulling real runtime state out of a running service without breaking it, were the hard parts we learnt to put together.
We learnt some other lessons the hard way too. HyperTest tried to prevent bugs with better tests, and adoption was a fight every time. Calls kept getting cancelled because teams were firefighting production. Testing was hygiene. Broken prod was hair on fire. This made us see where priorities lie.
This seeded the idea of building a truly autonomous on-call agent i.e. one that takes an alert, probes, diagnoses and fixes it in a few minutes. But this is how it works as of now:
You talk to your coding agent the way you already do. Tell it what's wrong: "checkout returns 200 but some users are seeing their order fail, find out why." It locates the line in your local code, connects to us over MCP, and drops a probe on that line in the running service. The probe is read-only and sits dormant until real traffic hits. When hit, it captures the local variables at every frame of the call stack at that exact moment. It hands them to the agent, which diagnoses with real data.
There are two pieces. An SDK that runs inside your service, and an MCP server your coding agent talks to. The SDK is what makes setting probes (virtual breakpoints, log or metric) possible without a redeploy. In Node and Python it hooks in-process. In Java it attaches as a JVM agent, instrumenting at the bytecode level. Either way the service keeps running and serving traffic. Nothing pauses.
When your agent wants to look at a line, it calls the MCP server, which tells the SDK to place a probe there. When a request hits the line, the SDK captures what the probe asked for, sanitizes it in-process, and streams it back to the agent via the MCP.
This can run inside prod, so a probe can read any value sitting in that variable. We ensure redaction happens in-process, inside your own container's memory. This is before anything goes on the wire. Keys like password, token, authorization, ssn and credit card are redacted by default and you add your own.
Also the probes read but never write, and if you want no captured state to ever leave your network you can self host the server, broker, and even the database in your infra.
On overhead: when idle, the SDK adds negligible memory and effectively nothing to throughput and response time. Probes only cost anything while actively capturing. Also captures are bounded. A separate monitor watches in real time and pulls every active probe if overhead ever spikes.
Every log-and-trace tool hands the agent data that already exists and asks it to reason backward to what probably happened. We think it is more useful to give agents eyes and ears into the running code, so they capture what they need when they need it, right at the point of failure.
This seems like the simplest and fastest way to debug prod incidents.
We\u2019d love the community to try this in any environment to debug any known or unknown issue by just chatting with your coding agent. And let us know what more features you need to make this a truly autonomous on-call agent
Supported platforms: NodeJs, Java, Python.","title":"Launch HN: HyperProbe (YC S26) \u2013 Agents that do read-only debugging in prod","updated_at":"2026-09-08T01:37:00Z","url":"https://www.hyperprobe.co"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"merterdir"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"Hello HN!
Remember is an intelligent personal search engine that lets you find every moment of your digital life.
We connect to your computer, phone, wearables and the services you use (Facebook, Twitter, mobile apps, etc.) and make all of your data searchable from one place. Going a step further, we run machine learning algorithms on all of your files to understand context so you can search using natural language.
We lose so much time looking for our files; be it notes, messages, photos, videos, audio recordings.
"Let me show you a video I sent Jack it's hilarious!" Stares at phone for 15 minutes looking for it
"I have a breakthrough idea about mitosis, let me write it down." Never gets back to that note again
This must end. We want to help maximise human potential, and we think this as close as we can get to having perfect memory with today's technology.
Founders:
Mert Erdir - Experience Design: BA Product Design in Milan, Italy. Ex-brand ambassador for Samsung, Timbuk2, Betabrand.
Fun fact: http://gizmodo.com/5809621/how-to-upgrade-to-ios-5-today-wit...
Nisan Gulsar - Experience Design: Advertising student. Weteraanikonepaivat (Finland) and Sinfonia di Fiori (Italy) art festivals.
Yigit Demirag - AI Researcher: PH-SFT at CERN (Switzerland) , Google Summer of Code, Bilkent University Nanotechnology Research Center
Kerem Gunes - Back end dev: KAREL Electronics
We know that this is insanely hard to build. We're not saying it's a piece of cake for us, but we're planning to get there by keeping learning and working. If you would love to have instant access to all of your life's work and memories, vote for us!
Demo: We have a demo, but for technical reasons I can't just post it here yet. You can sign up to the newsletter on our website to be the first to know when the beta is out. If you can't wait, send me an email and I'll send you the demo link: mert@remember.ai
http://remember.ai"},"title":{"matchLevel":"none","matchedWords":[],"value":"Apply HN: Remember \u2013 Intelligently search all of your files from one place."}},"_tags":["story","author_merterdir","story_11483510","ask_hn"],"author":"merterdir","children":[11484020,11484087,11488028,11488616,11491676,11494553,11520840],"created_at":"2016-04-12T20:55:27Z","created_at_i":1460494527,"num_comments":12,"objectID":"11483510","points":18,"story_id":11483510,"story_text":"Hello HN!
Remember is an intelligent personal search engine that lets you find every moment of your digital life.
We connect to your computer, phone, wearables and the services you use (Facebook, Twitter, mobile apps, etc.) and make all of your data searchable from one place. Going a step further, we run machine learning algorithms on all of your files to understand context so you can search using natural language.
We lose so much time looking for our files; be it notes, messages, photos, videos, audio recordings.
"Let me show you a video I sent Jack it's hilarious!" Stares at phone for 15 minutes looking for it
"I have a breakthrough idea about mitosis, let me write it down." Never gets back to that note again
This must end. We want to help maximise human potential, and we think this as close as we can get to having perfect memory with today's technology.
Founders:
Mert Erdir - Experience Design: BA Product Design in Milan, Italy. Ex-brand ambassador for Samsung, Timbuk2, Betabrand.
Fun fact: http://gizmodo.com/5809621/how-to-upgrade-to-ios-5-today-wit...
Nisan Gulsar - Experience Design: Advertising student. Weteraanikonepaivat (Finland) and Sinfonia di Fiori (Italy) art festivals.
Yigit Demirag - AI Researcher: PH-SFT at CERN (Switzerland) , Google Summer of Code, Bilkent University Nanotechnology Research Center
Kerem Gunes - Back end dev: KAREL Electronics
We know that this is insanely hard to build. We're not saying it's a piece of cake for us, but we're planning to get there by keeping learning and working. If you would love to have instant access to all of your life's work and memories, vote for us!
Demo: We have a demo, but for technical reasons I can't just post it here yet. You can sign up to the newsletter on our website to be the first to know when the beta is out. If you can't wait, send me an email and I'll send you the demo link: mert@remember.ai
http://remember.ai","title":"Apply HN: Remember \u2013 Intelligently search all of your files from one place.","updated_at":"2024-09-19T23:06:56Z"},{"_highlightResult":{"author":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["karen"],"value":"mr-karan"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"Hey HN, I'm Karan. I built Logchef because we moved our log workloads to ClickHouse at Zerodha and found the ecosystem lacking for actual log exploration.
Tools like Metabase and Grafana are great for dashboards, but log exploration is a different workflow.
The Grafana ClickHouse plugin requires manual column configuration for logs, has no field sidebar, and its ad-hoc filters are slow on large tables.\nGrafana's RBAC is Enterprise-only, and even then it's datasource-level and not at table-level. If someone has access to the ClickHouse datasource, they can query any table. In Logchef, you create Sources (pointing to specific tables) and assign them to Teams. Team A only sees their logs, Team B only theirs. Fine-grained access control without any Enterprise license :)
Metabase has no concept of log-specific features at all. We kept falling back to raw SQL in clickhouse-client, which works but isn't something you want to hand to your whole team.
So I built Logchef \u2014 a log viewer designed specifically for ClickHouse. Schema-agnostic, so you can start filtering immediately without migrations.
What it does:
- Point at any ClickHouse table \u2013 no schema migrations/no OTEL requirement\n- LogchefQL query language (simpler than SQL for filtering)\n- Field sidebar like Kibana \u2013 click values to add filters\n- SQL-based alerting that routes to Alertmanager\n- AI assistant that turns natural language queries like "find slow POST requests" into working SQL\n- MCP server for querying logs from your favourite LLM assistant tool\n- Single Go binary (~15MB), embeds the frontend
We've been running this in production for 8 months and I'm excited to ship v1.0 today.
Demo: https://demo.logchef.app
GitHub: https://github.com/mr-karan/logchef
Docs: https://logchef.app
Happy to answer questions about ClickHouse for logs, the architecture, or anything else."},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Logchef \u2013 Open-source schema-agnostic log viewer for ClickHouse"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://logchef.app/"}},"_tags":["story","author_mr-karan","story_46356369","show_hn"],"author":"mr-karan","children":[46356887,46419650],"created_at":"2025-12-22T17:35:46Z","created_at_i":1766424946,"num_comments":3,"objectID":"46356369","points":5,"story_id":46356369,"story_text":"Hey HN, I'm Karan. I built Logchef because we moved our log workloads to ClickHouse at Zerodha and found the ecosystem lacking for actual log exploration.
Tools like Metabase and Grafana are great for dashboards, but log exploration is a different workflow.
The Grafana ClickHouse plugin requires manual column configuration for logs, has no field sidebar, and its ad-hoc filters are slow on large tables.\nGrafana's RBAC is Enterprise-only, and even then it's datasource-level and not at table-level. If someone has access to the ClickHouse datasource, they can query any table. In Logchef, you create Sources (pointing to specific tables) and assign them to Teams. Team A only sees their logs, Team B only theirs. Fine-grained access control without any Enterprise license :)
Metabase has no concept of log-specific features at all. We kept falling back to raw SQL in clickhouse-client, which works but isn't something you want to hand to your whole team.
So I built Logchef \u2014 a log viewer designed specifically for ClickHouse. Schema-agnostic, so you can start filtering immediately without migrations.
What it does:
- Point at any ClickHouse table \u2013 no schema migrations/no OTEL requirement\n- LogchefQL query language (simpler than SQL for filtering)\n- Field sidebar like Kibana \u2013 click values to add filters\n- SQL-based alerting that routes to Alertmanager\n- AI assistant that turns natural language queries like "find slow POST requests" into working SQL\n- MCP server for querying logs from your favourite LLM assistant tool\n- Single Go binary (~15MB), embeds the frontend
We've been running this in production for 8 months and I'm excited to ship v1.0 today.
Demo: https://demo.logchef.app
GitHub: https://github.com/mr-karan/logchef
Docs: https://logchef.app
Happy to answer questions about ClickHouse for logs, the architecture, or anything else.","title":"Show HN: Logchef \u2013 Open-source schema-agnostic log viewer for ClickHouse","updated_at":"2026-03-05T23:16:26Z","url":"https://logchef.app/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"soham123"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"Hey HN,
I built SWE-Kit, LLM toolkit (Function callable tools) which makes building agents specialised in coding like Devin very easy.
I noticed a typical pattern while building local agents: creating & perfecting LLM tools to interact with system or codebase was the repeated and time-consuming. We created a layer that simplifies building agents that can interact with code, file system, git, shell and allows you to quickly solve for a wide variety of coding agent use cases.
Aren\u2019t there open coding agents already? Well, yes, but most folks would want to solve their specific use case like a large refactor and current coding agents aren\u2019t customisable to your specific use case or aren\u2019t meant to be molded to different workflows.
The idea is to provide a library of tools so you can build software engineering agents with a few lines of code in agentic framework of your choice.
We have solved following hard parts for everyone - \n- Optimized Coding Tools: Includes Code Analysis, File Operations, and Shell tools for seamless interaction with codebases and operating systems.\n- Browser Interaction Tool: Enables navigation and interaction with UI-based applications and codebases.\n- Framework Agnostic: Compatible with frameworks like LangChain, LlamaIndex, CrewAI, and Autogen, this allows you to work with your preferred setup.\n- Third-Party Integrations: Connects with applications like GitHub, Slack, Jira, and Gmail to build fully autonomous, end-to-end AI coding agents.\n- Flexible Deployment: Run on Local, Docker, Fly.io, E2b, AWS Lambda (soon!)
Is this the 10x Coding Agent I was looking for?
No this is not a coding agent but allows you to build your custom coding agent in framework of your choice.
We have created some templates to get started quickly though:\n- GitHub PR Agent: Autonomously reviews GitHub pull requests with full codebase context.\n- SWE Agent: Writes new features, debugs code, refactors, and creates tests.\n- Codebase Q&A Agent: Enables natural language interactions with the codebase.
To better showcase the SWE kit's capability, we tested it on [swebench.com](https://www.swebench.com/), the benchmark for testing coding agents. It scored 48.60%, whereas Devin scored only 13.86%.
If you end up using this, please do provide feedback and if you need help building coding agent feel free to reach out to us
I (Soham) & my cofounder Karan are both active on this thread to answer any questions!"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: LLM Function Calling Library to Interact with File, Shell, Git and Code"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://swekit.dev/"}},"_tags":["story","author_soham123","story_42063098","show_hn"],"author":"soham123","created_at":"2024-11-06T14:47:44Z","created_at_i":1730904464,"num_comments":0,"objectID":"42063098","points":5,"story_id":42063098,"story_text":"Hey HN,
I built SWE-Kit, LLM toolkit (Function callable tools) which makes building agents specialised in coding like Devin very easy.
I noticed a typical pattern while building local agents: creating & perfecting LLM tools to interact with system or codebase was the repeated and time-consuming. We created a layer that simplifies building agents that can interact with code, file system, git, shell and allows you to quickly solve for a wide variety of coding agent use cases.
Aren\u2019t there open coding agents already? Well, yes, but most folks would want to solve their specific use case like a large refactor and current coding agents aren\u2019t customisable to your specific use case or aren\u2019t meant to be molded to different workflows.
The idea is to provide a library of tools so you can build software engineering agents with a few lines of code in agentic framework of your choice.
We have solved following hard parts for everyone - \n- Optimized Coding Tools: Includes Code Analysis, File Operations, and Shell tools for seamless interaction with codebases and operating systems.\n- Browser Interaction Tool: Enables navigation and interaction with UI-based applications and codebases.\n- Framework Agnostic: Compatible with frameworks like LangChain, LlamaIndex, CrewAI, and Autogen, this allows you to work with your preferred setup.\n- Third-Party Integrations: Connects with applications like GitHub, Slack, Jira, and Gmail to build fully autonomous, end-to-end AI coding agents.\n- Flexible Deployment: Run on Local, Docker, Fly.io, E2b, AWS Lambda (soon!)
Is this the 10x Coding Agent I was looking for?
No this is not a coding agent but allows you to build your custom coding agent in framework of your choice.
We have created some templates to get started quickly though:\n- GitHub PR Agent: Autonomously reviews GitHub pull requests with full codebase context.\n- SWE Agent: Writes new features, debugs code, refactors, and creates tests.\n- Codebase Q&A Agent: Enables natural language interactions with the codebase.
To better showcase the SWE kit's capability, we tested it on [swebench.com](https://www.swebench.com/), the benchmark for testing coding agents. It scored 48.60%, whereas Devin scored only 13.86%.
If you end up using this, please do provide feedback and if you need help building coding agent feel free to reach out to us
I (Soham) & my cofounder Karan are both active on this thread to answer any questions!","title":"Show HN: LLM Function Calling Library to Interact with File, Shell, Git and Code","updated_at":"2024-11-18T04:36:30Z","url":"https://swekit.dev/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"karanch"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"Coming soon to an iPhone near you
That sweet hypnotic robotic voice that you hear from an iPhone 4S nowadays might not sound that sweet to some people after all. For beginners, definitely not to Google. No, its not the Android factor, in fact what seems to be worrying the search giant is the ad-stripped search results that Siri is dishing out to its users.
What might seem like a cool gadgetry app or just a fun activity today \u2013 that is, talking to your phone or making it answer your silly questions, is potentially as destructive for its competition as its innovative for the consumers. Siri in more ways than one has the potential to redefine the way we perceive and interact with a plethora of services that are available online. And in doing so, its paving the way for people to make these services a part of their everyday life closer than ever before.
The web has already become a cohesive unit of services with every application trying to find their niche market. You have Netflix for movies, Facebook for social connect, Spotify to listen to your music, Yelp to find your local grub and so on. Siri currently uses close to 35 API\u2019s for retrieving information which covers most of the basic services that should satisfy most of the users needs. But this number is surely going to increase and as the app space becomes more cramped, and its those that find a niche, will be the ones who will survive.
That begs a very important question? What API\u2019s are going to be used by Siri in the future.
For example, Bing would surely feel short changed, as Siri uses google as the default search engine. Not that you cant change this in iphone settings, the question is how many users will do it? The same goes with yelp vs open table or a rotten tomato vs metacritic and the list goes on.
The interesting thing to watch would be if Siri will ever release its own API or just continue to partner with a host of services as they continue to move ahead. Also, until now we have seen Siri as an interface to collect data from a source, will she ever be used to push data to a service?
I am sure a lot of developers are waiting with bated breath for the day Siri releases its API. For the possibility of integrating with this AI machine only means getting more people and traffic towards your service. Google surely is not going to wait and watch. Iris, the rumored Siri alternative for android is on its way and will surely try and match Siri blow by blow.
Looking at the history of the two companies, Google has always been perceived as the developer friendly one, and the assumption its Siri\u2019s alternative will have its API released should not be a long shot. On the other hand, Apple has always played daddy to with its rules and restrictions. What say Flash? Will it change its game this time, considering it already has an edge with the hype it has generated with Siri.
Finally, I guess if you are an app developer or a web service provider, the most important question posing you tomorrow would be \u2013 Are you on Siri, yet? Cause maybe you might just not want to miss this boat.
Deepak Venugopal and Karan Chaudhry"},"title":{"matchLevel":"none","matchedWords":[],"value":"Siri - The battle of the API\u2019s "},"url":{"matchLevel":"none","matchedWords":[],"value":""}},"_tags":["story","author_karanch","story_3234931","ask_hn"],"author":"karanch","created_at":"2011-11-14T17:58:56Z","created_at_i":1321293536,"num_comments":0,"objectID":"3234931","points":1,"story_id":3234931,"story_text":"Coming soon to an iPhone near you
That sweet hypnotic robotic voice that you hear from an iPhone 4S nowadays might not sound that sweet to some people after all. For beginners, definitely not to Google. No, its not the Android factor, in fact what seems to be worrying the search giant is the ad-stripped search results that Siri is dishing out to its users.
What might seem like a cool gadgetry app or just a fun activity today \u2013 that is, talking to your phone or making it answer your silly questions, is potentially as destructive for its competition as its innovative for the consumers. Siri in more ways than one has the potential to redefine the way we perceive and interact with a plethora of services that are available online. And in doing so, its paving the way for people to make these services a part of their everyday life closer than ever before.
The web has already become a cohesive unit of services with every application trying to find their niche market. You have Netflix for movies, Facebook for social connect, Spotify to listen to your music, Yelp to find your local grub and so on. Siri currently uses close to 35 API\u2019s for retrieving information which covers most of the basic services that should satisfy most of the users needs. But this number is surely going to increase and as the app space becomes more cramped, and its those that find a niche, will be the ones who will survive.
That begs a very important question? What API\u2019s are going to be used by Siri in the future.
For example, Bing would surely feel short changed, as Siri uses google as the default search engine. Not that you cant change this in iphone settings, the question is how many users will do it? The same goes with yelp vs open table or a rotten tomato vs metacritic and the list goes on.
The interesting thing to watch would be if Siri will ever release its own API or just continue to partner with a host of services as they continue to move ahead. Also, until now we have seen Siri as an interface to collect data from a source, will she ever be used to push data to a service?
I am sure a lot of developers are waiting with bated breath for the day Siri releases its API. For the possibility of integrating with this AI machine only means getting more people and traffic towards your service. Google surely is not going to wait and watch. Iris, the rumored Siri alternative for android is on its way and will surely try and match Siri blow by blow.
Looking at the history of the two companies, Google has always been perceived as the developer friendly one, and the assumption its Siri\u2019s alternative will have its API released should not be a long shot. On the other hand, Apple has always played daddy to with its rules and restrictions. What say Flash? Will it change its game this time, considering it already has an edge with the hype it has generated with Siri.
Finally, I guess if you are an app developer or a web service provider, the most important question posing you tomorrow would be \u2013 Are you on Siri, yet? Cause maybe you might just not want to miss this boat.
Deepak Venugopal and Karan Chaudhry","title":"Siri - The battle of the API\u2019s ","updated_at":"2024-09-19T18:04:45Z","url":""},{"_highlightResult":{"author":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["karen"],"value":"karan_dev"},"title":{"matchLevel":"none","matchedWords":[],"value":"RageDetector \u2013 detects aggressive typing and forces me to calm down"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"https://github.com/AI-Architechs/RageDetector"}},"_tags":["story","author_karan_dev","story_47057254"],"author":"karan_dev","children":[47057255,47057646],"created_at":"2026-02-18T04:44:33Z","created_at_i":1771389873,"num_comments":2,"objectID":"47057254","points":3,"story_id":47057254,"title":"RageDetector \u2013 detects aggressive typing and forces me to calm down","updated_at":"2026-03-05T23:34:56Z","url":"https://github.com/AI-Architechs/RageDetector"},{"_highlightResult":{"author":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["karen"],"value":"Suyash_karn"},"title":{"matchLevel":"none","matchedWords":[],"value":"Static websites had a good run"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"https://www.interactlabs.ai/blog-article/introducing-interact-ai"}},"_tags":["story","author_Suyash_karn","story_48184139"],"author":"Suyash_karn","created_at":"2026-05-18T19:08:59Z","created_at_i":1779131339,"num_comments":0,"objectID":"48184139","points":2,"story_id":48184139,"title":"Static websites had a good run","updated_at":"2026-05-19T09:18:27Z","url":"https://www.interactlabs.ai/blog-article/introducing-interact-ai"},{"_highlightResult":{"author":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["karen"],"value":"Suyash_karn"},"title":{"matchLevel":"none","matchedWords":[],"value":"See how websites might look like in the future"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"https://www.interactlabs.ai/?auto_open=true"}},"_tags":["story","author_Suyash_karn","story_48184294"],"author":"Suyash_karn","children":[48184295,48190294],"created_at":"2026-05-18T19:26:11Z","created_at_i":1779132371,"num_comments":2,"objectID":"48184294","points":1,"story_id":48184294,"title":"See how websites might look like in the future","updated_at":"2026-05-19T11:48:13Z","url":"https://www.interactlabs.ai/?auto_open=true"},{"_highlightResult":{"author":{"fullyHighlighted":true,"matchLevel":"partial","matchedWords":["karen"],"value":"karenL"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"matchLevel":"none","matchedWords":[],"value":"Balanced Surface RT Review..."},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"http://www.airbornegeek.com/2012/10/microsoft-surface-rt-how-is-it/"}},"_tags":["story","author_karenL","story_4719033"],"author":"karenL","created_at":"2012-10-30T19:25:25Z","created_at_i":1351625125,"num_comments":0,"objectID":"4719033","points":1,"story_id":4719033,"story_text":"","title":"Balanced Surface RT Review...","updated_at":"2024-09-19T18:56:00Z","url":"http://www.airbornegeek.com/2012/10/microsoft-surface-rt-how-is-it/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"ingve"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["karen","ai"],"value":"Why aren't AI companies being charged for hacking?"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai"],"value":"https://mortoray.com/why-arent-ai-companies-being-charged-for-hacking/"}},"_tags":["story","author_ingve","story_50057278"],"author":"ingve","children":[50057329,50057331,50057342,50057343,50057347,50057348,50057353,50057360,50057381,50057392,50057406,50057432,50057441,50057452,50057467,50057481,50057500,50057501,50057515,50057530,50057543,50057545,50057557,50057562,50057575,50057585,50057624,50057631,50057672,50057699,50057775,50057794,50057942,50057989,50058025,50058085,50058098,50058132,50058183,50058228,50058365,50058444,50058685,50058724,50058792,50058975,50059267,50060157,50061849,50062339,50062869,50064180,50066141,50068548,50069066],"created_at":"2026-10-12T14:52:37Z","created_at_i":1791816757,"num_comments":372,"objectID":"50057278","points":336,"story_id":50057278,"title":"Why aren't AI companies being charged for hacking?","updated_at":"2026-10-16T04:27:02Z","url":"https://mortoray.com/why-arent-ai-companies-being-charged-for-hacking/"}],"hitsPerPage":50,"nbHits":1309,"nbPages":20,"page":0,"params":"query=Karen+AI&tags=story&hitsPerPage=50&advancedSyntax=true&analyticsTags=backend","processingTimeMS":24,"processingTimingsMS":{"_request":{"queue":115,"roundTrip":22},"afterFetch":{"format":{"highlighting":3,"total":3},"merge":{"mergeLoop":{"prepareNextHit":1,"total":1},"total":1},"total":1},"fetch":{"query":11,"scanning":10,"total":22},"total":24},"query":"Karen AI","serverTimeMS":143}