Traditional code search tools match the terms in your query against the codebase, but often you don\u2019t know the right terms to start with, e.g. \u2018Which library do we use for model inference?\u2019 (These types of questions are particularly common when you\u2019re learning a new codebase.) bloop uses a combination of neural semantic code search (comparing the meaning - encoded in vector representations - of queries and code snippets) and chained LLM calls to retrieve and reason about abstract queries.
Ideally, a LLM could answer questions about your code directly, but there is significant overhead (and expense) in fine-tuning the largest LLMs on private data. And although they\u2019re increasing, prompt sizes are still a long way off being able to fit a whole organisation\u2019s codebase.
We get around these limitations with a two-step process. First, we use GPT-4 to generate a keyword query which is passed to a semantic search engine. This embeds the query and compares it to chunks of code in vector space (we use Qdrant as our vector DB). We\u2019ve found that using a semantic search engine for retrieval improves recall, allowing the LLM to retrieve code that doesn\u2019t have any textual overlap with the query but is still relevant. Second, the retrieved code snippets are ranked and inserted into a final LLM prompt. We pass this to GPT-4 and its phenomenal understanding of code does the rest.
Let\u2019s work through an example. You start off by asking \u2018Where is the query parsing logic?\u2019 and then want to find out \u2018Which library does it use?\u2019. We use GPT-4 to generate the standalone keyword query: \u2018query parser library\u2019, which we then pass to a semantic search engine that returns a snippet demonstrating the parser in action: \u2018let pair = PestParser::parse(Rule::query, query);\u2019. We insert this snippet into a prompt to GPT-4, which is able to work out that pest is the library doing the legwork here, generating the answer \u2018The query parser uses the pest library\u2019.
You can also filter your search by repo or language - What\u2019s the readiness delay repo:myApp lang:yaml. GPT-4 will generate an answer constrained to the respective repo and language.
We also know that LLMs are not always (at least not yet) the best tool for the job. Sometimes you know exactly what you\u2019re looking for. For this, we\u2019ve built a fast, trigram index based regex search engine based on Tantivy. Because of this, bloop is fast at traditional search too. For code navigation, we\u2019ve built a precise go-to-ref/def engine based on scope resolution that uses Tree-sitter.
bloop is fully open-source. Semantic search, LLM prompts, regex search and code navigation are all contained in one repo: https://github.com/bloopAI/bloop.
Our software is standalone and doesn\u2019t run in your IDE. We were originally IDE-based but moved away from this due to constraints on how we could display code to the user.
bloop runs as a free desktop app on Mac, Windows and Linux: https://github.com/bloopAI/bloop/releases. On desktop, your code is indexed with a MiniLM embedding model and stored locally, meaning at index time your codebase stays private. Indexing is fast, except on the very largest repos (GPU indexing coming soon). \u2018Private\u2019 here means that no code is shared with us or OpenAI at index time, and when a search is made only relevant code snippets are shared to generate the response. (This is more or less the same data usage as Copilot).
We also have a paid cloud offering for teams ($12 per user per month). Members of the same organisation can search a shared index hosted by us.
We\u2019d love to hear your thoughts about the product and where you think we should take it next, and your thoughts on code search in general. We look forward to your comments!"},"title":{"matchLevel":"none","matchedWords":[],"value":"Launch HN: Bloop (YC S21) \u2013 Code Search with GPT-4"}},"_tags":["story","author_louiskw","story_35236275","launch_hn"],"author":"louiskw","children":[35236557,35236562,35236580,35236687,35236748,35236752,35236968,35237012,35237062,35237083,35237120,35237496,35237781,35237851,35237865,35238126,35238292,35238325,35238628,35241680,35242030,35242293,35242546,35243638,35245284,35252258],"created_at":"2023-03-20T18:27:32Z","created_at_i":1679336852,"num_comments":125,"objectID":"35236275","points":264,"story_id":35236275,"story_text":"Hey HN, we\u2019re the cofounders of bloop (https://bloop.ai/), a code search engine which combines semantic search with GPT-4 to answer questions. We let you search your private codebases either the traditional way (regex or literal) or via semantic search, ask questions in natural language thanks to GPT-4, and jump between refs/defs with precise code navigation. Here\u2019s a quick demo: https://www.loom.com/share/8e9d59b88dd2409482ec02cdda5b9185
Traditional code search tools match the terms in your query against the codebase, but often you don\u2019t know the right terms to start with, e.g. \u2018Which library do we use for model inference?\u2019 (These types of questions are particularly common when you\u2019re learning a new codebase.) bloop uses a combination of neural semantic code search (comparing the meaning - encoded in vector representations - of queries and code snippets) and chained LLM calls to retrieve and reason about abstract queries.
Ideally, a LLM could answer questions about your code directly, but there is significant overhead (and expense) in fine-tuning the largest LLMs on private data. And although they\u2019re increasing, prompt sizes are still a long way off being able to fit a whole organisation\u2019s codebase.
We get around these limitations with a two-step process. First, we use GPT-4 to generate a keyword query which is passed to a semantic search engine. This embeds the query and compares it to chunks of code in vector space (we use Qdrant as our vector DB). We\u2019ve found that using a semantic search engine for retrieval improves recall, allowing the LLM to retrieve code that doesn\u2019t have any textual overlap with the query but is still relevant. Second, the retrieved code snippets are ranked and inserted into a final LLM prompt. We pass this to GPT-4 and its phenomenal understanding of code does the rest.
Let\u2019s work through an example. You start off by asking \u2018Where is the query parsing logic?\u2019 and then want to find out \u2018Which library does it use?\u2019. We use GPT-4 to generate the standalone keyword query: \u2018query parser library\u2019, which we then pass to a semantic search engine that returns a snippet demonstrating the parser in action: \u2018let pair = PestParser::parse(Rule::query, query);\u2019. We insert this snippet into a prompt to GPT-4, which is able to work out that pest is the library doing the legwork here, generating the answer \u2018The query parser uses the pest library\u2019.
You can also filter your search by repo or language - What\u2019s the readiness delay repo:myApp lang:yaml. GPT-4 will generate an answer constrained to the respective repo and language.
We also know that LLMs are not always (at least not yet) the best tool for the job. Sometimes you know exactly what you\u2019re looking for. For this, we\u2019ve built a fast, trigram index based regex search engine based on Tantivy. Because of this, bloop is fast at traditional search too. For code navigation, we\u2019ve built a precise go-to-ref/def engine based on scope resolution that uses Tree-sitter.
bloop is fully open-source. Semantic search, LLM prompts, regex search and code navigation are all contained in one repo: https://github.com/bloopAI/bloop.
Our software is standalone and doesn\u2019t run in your IDE. We were originally IDE-based but moved away from this due to constraints on how we could display code to the user.
bloop runs as a free desktop app on Mac, Windows and Linux: https://github.com/bloopAI/bloop/releases. On desktop, your code is indexed with a MiniLM embedding model and stored locally, meaning at index time your codebase stays private. Indexing is fast, except on the very largest repos (GPU indexing coming soon). \u2018Private\u2019 here means that no code is shared with us or OpenAI at index time, and when a search is made only relevant code snippets are shared to generate the response. (This is more or less the same data usage as Copilot).
We also have a paid cloud offering for teams ($12 per user per month). Members of the same organisation can search a shared index hosted by us.
We\u2019d love to hear your thoughts about the product and where you think we should take it next, and your thoughts on code search in general. We look forward to your comments!","title":"Launch HN: Bloop (YC S21) \u2013 Code Search with GPT-4","updated_at":"2024-09-20T13:36:55Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"matijash"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["andreason"],"value":"Hi HN!
We are Martin and Matija, twin brothers and creators of Wasp (https://wasp-lang.dev). Wasp is a declarative language that makes it really easy to build full-stack web apps while still using the latest technologies such as React, Node.js and Prisma.
Martin and I both studied computer science where we mostly focused on algorithms for bioinformatics. Afterwards we led engineering teams in several SaaS companies, on the way gaining plenty of experience in building web apps.
Moving from one project to another, we used various technologies:\nJQuery -> Backbone -> Angular -> React, own scripts / makefile -> Grunt -> Gulp -> Webpack, PHP -> Java -> Node.js, \u2026 , and we always felt that things are harder than they should be. We were spending a lot of time adopting the latest tech stack and figuring out the best practices: how to make the web app performant, scalable, economical and secure and also how to connect all the pieces of the stack together.
While the tech stack kept advancing rapidly, the core requirements of the apps we were building changed very little (auth, routing, data model CRUD, ACL, \u2026). That is why about 1.5 years ago we started thinking about separating web app specification (what it should do) from its implementation (how it should do it).
This led us to the idea of extracting common web app features and concepts into a special specification language from which we could generate code in the currently popular technologies. We don\u2019t think it is feasible to replace everything with a single language so that is why we went with a DSL which integrates with the modern stack (right now React, NodeJS, Prisma).
Wasp lets you define high-level aspects of your web app (auth, routing, ACL, data models, CRUD) via a simple specification language and then write your specific logic in React and Node.js. The majority of the code is still being written in React and Node.js, with Wasp serving as the backbone of your whole application. To see some examples of what the language looks like in practice, take a look here: https://github.com/wasp-lang/wasp/blob/master/examples/tutor...
The main difference between Wasp and frameworks (e.g. Meteor, Blitz, Redwood) is that Wasp is a language, not a library. One benefit of that is a simpler and cleaner, declarative syntax, focused on the requirements and detached from the implementation details.
Another benefit of a DSL is that it allows Wasp to understand the web app\u2019s requirements during the build time and reason about it before generating the final code. For example, when generating code to be deployed to production, it could pick the most appropriate architecture based on its understanding of the web app and deploy it to serverless or another type of architecture (or even a combination). Another example would be reusing your data model logic through all the parts of the stack while defining it just once in Wasp. DSL opens the potential for optimisations, static analysis and extensibility.
Wasp\u2019s compiler is built in Haskell and it compiles the source code in Wasp + React/Node.js into the target code in just React and Node.js (currently in Javascript, but we plan to move to Typescript soon). The generated code is human readable and can easily be inspected and even ejected if Wasp becomes too limiting.
We are currently in Alpha and many features are still rough or missing, but you can try it out and build and deploy web apps! There are things we haven\u2019t solved yet and others that will probably change as we progress.
You can check out our repo at https://github.com/wasp-lang/wasp and give it a try at https://wasp-lang.dev/docs/.
Thank you for reading! We would love to get your feedback and also hear about your experiences building web apps - what has worked for you and where do you see the opportunities for improvement?"},"title":{"matchLevel":"none","matchedWords":[],"value":"Launch HN: Wasp (YC W21) \u2013 DSL for building full-stack web apps"}},"_tags":["story","author_matijash","story_26091956","launch_hn"],"author":"matijash","children":[26092049,26092090,26092236,26092322,26092336,26092380,26092635,26092660,26092681,26092752,26092769,26092877,26092988,26093043,26093186,26093283,26093801,26094118,26094265,26094386,26094506,26096285,26097817,26100199,26100223,26102170,26102320,26107414],"created_at":"2021-02-10T17:15:50Z","created_at_i":1612977350,"num_comments":79,"objectID":"26091956","points":222,"story_id":26091956,"story_text":"Hi HN!
We are Martin and Matija, twin brothers and creators of Wasp (https://wasp-lang.dev). Wasp is a declarative language that makes it really easy to build full-stack web apps while still using the latest technologies such as React, Node.js and Prisma.
Martin and I both studied computer science where we mostly focused on algorithms for bioinformatics. Afterwards we led engineering teams in several SaaS companies, on the way gaining plenty of experience in building web apps.
Moving from one project to another, we used various technologies:\nJQuery -> Backbone -> Angular -> React, own scripts / makefile -> Grunt -> Gulp -> Webpack, PHP -> Java -> Node.js, \u2026 , and we always felt that things are harder than they should be. We were spending a lot of time adopting the latest tech stack and figuring out the best practices: how to make the web app performant, scalable, economical and secure and also how to connect all the pieces of the stack together.
While the tech stack kept advancing rapidly, the core requirements of the apps we were building changed very little (auth, routing, data model CRUD, ACL, \u2026). That is why about 1.5 years ago we started thinking about separating web app specification (what it should do) from its implementation (how it should do it).
This led us to the idea of extracting common web app features and concepts into a special specification language from which we could generate code in the currently popular technologies. We don\u2019t think it is feasible to replace everything with a single language so that is why we went with a DSL which integrates with the modern stack (right now React, NodeJS, Prisma).
Wasp lets you define high-level aspects of your web app (auth, routing, ACL, data models, CRUD) via a simple specification language and then write your specific logic in React and Node.js. The majority of the code is still being written in React and Node.js, with Wasp serving as the backbone of your whole application. To see some examples of what the language looks like in practice, take a look here: https://github.com/wasp-lang/wasp/blob/master/examples/tutor...
The main difference between Wasp and frameworks (e.g. Meteor, Blitz, Redwood) is that Wasp is a language, not a library. One benefit of that is a simpler and cleaner, declarative syntax, focused on the requirements and detached from the implementation details.
Another benefit of a DSL is that it allows Wasp to understand the web app\u2019s requirements during the build time and reason about it before generating the final code. For example, when generating code to be deployed to production, it could pick the most appropriate architecture based on its understanding of the web app and deploy it to serverless or another type of architecture (or even a combination). Another example would be reusing your data model logic through all the parts of the stack while defining it just once in Wasp. DSL opens the potential for optimisations, static analysis and extensibility.
Wasp\u2019s compiler is built in Haskell and it compiles the source code in Wasp + React/Node.js into the target code in just React and Node.js (currently in Javascript, but we plan to move to Typescript soon). The generated code is human readable and can easily be inspected and even ejected if Wasp becomes too limiting.
We are currently in Alpha and many features are still rough or missing, but you can try it out and build and deploy web apps! There are things we haven\u2019t solved yet and others that will probably change as we progress.
You can check out our repo at https://github.com/wasp-lang/wasp and give it a try at https://wasp-lang.dev/docs/.
Thank you for reading! We would love to get your feedback and also hear about your experiences building web apps - what has worked for you and where do you see the opportunities for improvement?","title":"Launch HN: Wasp (YC W21) \u2013 DSL for building full-stack web apps","updated_at":"2026-05-13T17:13:53Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"hsikka"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["andreason"],"value":"Hey HN,
I am a graduate student doing ML research, and lately I've been thinking a lot about designing learning systems from the hardware through the software layers.
I have no experience with what is going at the processor levels, and I was wondering what prerequisite subjects or general curricula I should follow to learn and reason at these lower levels of abstraction.
To be clear, I'm doing this to build intuitions about new computational systems and how different chips, from ASIC to neuromorphic, may be designed.
Any resources or advice telling me I'm a fool is welcome!"},"title":{"matchLevel":"none","matchedWords":[],"value":"Ask HN: How to Self-Study Integrated Circuit Design?"}},"_tags":["story","author_hsikka","story_19890949","ask_hn"],"author":"hsikka","children":[19891117,19893447,19893875,19894412,19895605,19896024,19897059,19897283,19897698,19898030,19898132,19899501,19899861,19901273,19910562,19912343,19918930,19945473],"created_at":"2019-05-12T09:26:26Z","created_at_i":1557653186,"num_comments":39,"objectID":"19890949","points":178,"story_id":19890949,"story_text":"Hey HN,
I am a graduate student doing ML research, and lately I've been thinking a lot about designing learning systems from the hardware through the software layers.
I have no experience with what is going at the processor levels, and I was wondering what prerequisite subjects or general curricula I should follow to learn and reason at these lower levels of abstraction.
To be clear, I'm doing this to build intuitions about new computational systems and how different chips, from ASIC to neuromorphic, may be designed.
Any resources or advice telling me I'm a fool is welcome!","title":"Ask HN: How to Self-Study Integrated Circuit Design?","updated_at":"2026-02-23T19:26:40Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"segmenta"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["andreason"],"value":"Claude Code is great, but it\u2019s focused on coding. The missing piece is a native way to build and run custom background agents for non-code tasks. We built RowboatX as a CLI tool modeled after Claude Code that lets you do that. It uses the file system and unix tools to create and monitor background agents for everyday tasks, connect them to any MCP server for tools, and reason over their outputs.
Because RowboatX runs locally with shell access, the agents can install tools, execute code, and automate anything you could do in a terminal with your explicit permission. It works with any compatible LLM, including open-source ones.
Our repo is https://github.com/rowboatlabs/rowboat, and there\u2019s a demo video here: https://youtu.be/cyPBinQzicY
For example, you can connect RowboatX to the ElevenLabs MCP server and create a background workflow that produces a NotebookLM-style podcast every day from recent AI-agent papers on arXiv. Or you can connect it to Google Calendar and Exa Search to research meeting attendees and generate briefs before each event.
You can try these with: `npx @rowboatlabs/rowboatx`
We combined three simple ideas:
1. File system as state: Each agent\u2019s instruction, memory, logs, and data are just files on disk, grepable, diffable, and local. For instance, you can just run: grep -rl '"agent":"<agent-name>"' ~/.rowboat/runs to list every run for a particular workflow.
2. The supervisor agent: A Claude Code style agent that can create and run background agents. It predominantly uses Unix commands to monitor, update, and schedule agents. LLMs handle Unix tools better than backend APIs [1][2], so we leaned into that. It can also probe any MCP server and attach the tools to the agents.
3. Human-in-the-loop: Each background agent can emit a human_request message when needed (e.g. drafting a tricky email or installing a tool) that pauses execution and waits for input before continuing. The supervisor coordinates this.
I started my career over a decade ago building spam detection models at Twitter, spending a lot of my time in the terminal with Unix commands for data analysis [0] and Vowpal Wabbit for modeling. When Claude Code came along, it felt familiar and amazing to work with. But trying to use it beyond code always felt a bit forced. We built RowboatX to bring that same workflow to everyday tasks. It is Apache-2.0 licensed and easily extendable.
While there are many agent builders, running on the user's terminal enables unique use cases like computer and browser automation that cloud-based tools can't match. This power requires careful safety design. We implemented command-level allow/deny lists, with containerization coming next. We\u2019ve tried to design for safety from day one, but we\u2019d love to hear the community\u2019s perspective on what additional safeguards or approaches you\u2019d consider important here.
We\u2019re excited to share RowboatX with everyone here. We\u2019d love to hear your thoughts and welcome contributions!
\u2014
[0] https://web.stanford.edu/class/cs124/kwc-unix-for-poets.pdf\n[1] https://arxiv.org/pdf/2405.06807\n[2] https://arxiv.org/pdf/2501.10132"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: RowboatX \u2013 open-source Claude Code for everyday automations"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/rowboatlabs/rowboat"}},"_tags":["story","author_segmenta","story_45970338","show_hn"],"author":"segmenta","children":[45970783,45973010,45973188,45974987,45976894,45978513,45978603,45982189,46098480],"created_at":"2025-11-18T18:50:00Z","created_at_i":1763491800,"num_comments":42,"objectID":"45970338","points":131,"story_id":45970338,"story_text":"Claude Code is great, but it\u2019s focused on coding. The missing piece is a native way to build and run custom background agents for non-code tasks. We built RowboatX as a CLI tool modeled after Claude Code that lets you do that. It uses the file system and unix tools to create and monitor background agents for everyday tasks, connect them to any MCP server for tools, and reason over their outputs.
Because RowboatX runs locally with shell access, the agents can install tools, execute code, and automate anything you could do in a terminal with your explicit permission. It works with any compatible LLM, including open-source ones.
Our repo is https://github.com/rowboatlabs/rowboat, and there\u2019s a demo video here: https://youtu.be/cyPBinQzicY
For example, you can connect RowboatX to the ElevenLabs MCP server and create a background workflow that produces a NotebookLM-style podcast every day from recent AI-agent papers on arXiv. Or you can connect it to Google Calendar and Exa Search to research meeting attendees and generate briefs before each event.
You can try these with: `npx @rowboatlabs/rowboatx`
We combined three simple ideas:
1. File system as state: Each agent\u2019s instruction, memory, logs, and data are just files on disk, grepable, diffable, and local. For instance, you can just run: grep -rl '"agent":"<agent-name>"' ~/.rowboat/runs to list every run for a particular workflow.
2. The supervisor agent: A Claude Code style agent that can create and run background agents. It predominantly uses Unix commands to monitor, update, and schedule agents. LLMs handle Unix tools better than backend APIs [1][2], so we leaned into that. It can also probe any MCP server and attach the tools to the agents.
3. Human-in-the-loop: Each background agent can emit a human_request message when needed (e.g. drafting a tricky email or installing a tool) that pauses execution and waits for input before continuing. The supervisor coordinates this.
I started my career over a decade ago building spam detection models at Twitter, spending a lot of my time in the terminal with Unix commands for data analysis [0] and Vowpal Wabbit for modeling. When Claude Code came along, it felt familiar and amazing to work with. But trying to use it beyond code always felt a bit forced. We built RowboatX to bring that same workflow to everyday tasks. It is Apache-2.0 licensed and easily extendable.
While there are many agent builders, running on the user's terminal enables unique use cases like computer and browser automation that cloud-based tools can't match. This power requires careful safety design. We implemented command-level allow/deny lists, with containerization coming next. We\u2019ve tried to design for safety from day one, but we\u2019d love to hear the community\u2019s perspective on what additional safeguards or approaches you\u2019d consider important here.
We\u2019re excited to share RowboatX with everyone here. We\u2019d love to hear your thoughts and welcome contributions!
\u2014
[0] https://web.stanford.edu/class/cs124/kwc-unix-for-poets.pdf\n[1] https://arxiv.org/pdf/2405.06807\n[2] https://arxiv.org/pdf/2501.10132","title":"Show HN: RowboatX \u2013 open-source Claude Code for everyday automations","updated_at":"2026-05-08T04:06:31Z","url":"https://github.com/rowboatlabs/rowboat"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"d_man"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["andreason"],"value":"I work as a developer (Java full stack if it makes any difference) and live quite far from my workplace: it usually takes me one hour and a half to and from work by car. Moving closer to my workplace is not an option nor is it using other means to go back and forth (eg train/bus).\nI feel like my time spent on the road is mostly wasted: I'd like to find something IT-related to do (both listen to and reason about) that can help me learn new stuff. \nSince I'm passionate about Linux, I usually have a few Linux (or BSD) podcasts to listen to while driving but I found that to be not really instructive, just more enjoyable than listening to the radio or some music playlist. \nI'm looking for any kind of suggestion from fellow developers who are in the same spot and have found any interesting way or resources to 'spend wisely' their travel time. \nWhat resources would you suggest (if any exists) to learn something while driving (much like you would do with a foreign language audio course)? Is it even possible? Has anyone tried or can share an experience of 'audio learning' IT related?"},"title":{"matchLevel":"none","matchedWords":[],"value":"Ask HN: As a developer, how can I take advantage of time spent driving to work?"}},"_tags":["story","author_d_man","story_13000859","ask_hn"],"author":"d_man","children":[13000966,13001059,13001064,13001117,13001145,13001149,13001150,13001181,13001185,13001188,13001199,13001205,13001217,13001222,13001274,13001285,13001327,13001332,13001375,13001438,13001470,13001481,13001518,13001565,13001632,13001644,13001775,13002511,13002533,13002553,13002601,13002621,13002761,13002829,13003271,13003336,13003422,13003518,13003572,13004220,13004517,13005698,13008521,13009219,13011521,13011695,13013133,13014892,13016780,13029486],"created_at":"2016-11-20T20:05:19Z","created_at_i":1479672319,"num_comments":113,"objectID":"13000859","points":93,"story_id":13000859,"story_text":"I work as a developer (Java full stack if it makes any difference) and live quite far from my workplace: it usually takes me one hour and a half to and from work by car. Moving closer to my workplace is not an option nor is it using other means to go back and forth (eg train/bus).\nI feel like my time spent on the road is mostly wasted: I'd like to find something IT-related to do (both listen to and reason about) that can help me learn new stuff. \nSince I'm passionate about Linux, I usually have a few Linux (or BSD) podcasts to listen to while driving but I found that to be not really instructive, just more enjoyable than listening to the radio or some music playlist. \nI'm looking for any kind of suggestion from fellow developers who are in the same spot and have found any interesting way or resources to 'spend wisely' their travel time. \nWhat resources would you suggest (if any exists) to learn something while driving (much like you would do with a foreign language audio course)? Is it even possible? Has anyone tried or can share an experience of 'audio learning' IT related?","title":"Ask HN: As a developer, how can I take advantage of time spent driving to work?","updated_at":"2024-09-19T23:56:55Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"patelajay285"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["andreason"],"value":"Hi, this is Ajay and Alex, and we\u2019re the founders of Plasticity (https://www.plasticity.ai/). We're building an API that helps developers create human-like natural language interfaces.
Four years ago, we hacked 3rd party commands into Siri without jailbreaking before Alexa Skills or SiriKit were released (https://www.wired.com/2014/04/googolplex/). It was the first App Store for voice commands. Since then, we\u2019ve worked on NL interfaces at Google and Apple Siri. Now we're tackling the next problem: products using NLP are fairly simplistic in what they can do for users. For example, systems like Siri still struggle to directly answer a basic question like "When is the Y Combinator application due?" because it can't understand and reason where the answer may lie in a sentence on Y Combinator's website.
We\u2019re approaching the problem differently by understanding the structure of language and relationships within text, instead of relying on more simplistic methods like keyword matching. We build a graph of entities and their relationships within a sentence along with other linguistic information. You can think of it as \u201cOpen Information Extraction\u201d with a lot more information (https://www.plasticity.ai/api/demo).
Currently, we use a TensorFlow model to perform classical tasks like parts of speech, tokenization, and syntax dependency trees. We built our own Wikipedia crawler for data to better handle chunking and disambiguation, which helps return more accurate results for multi-word entities in sentences like: "The band played let it be by the beatles." We wrote our open IE algorithms from scratch, focusing on speed. It's written completely in C++ and we are adding more features everyday.
Our public APIs are in beta right now, we\u2019re constantly working to improve the accuracy, and we\u2019re looking forward to hearing feedback. We\u2019d love to hear what the HN community is working on with NLP and how we can help!"},"title":{"matchLevel":"none","matchedWords":[],"value":"Launch HN: Plasticity (YC S17) \u2013 APIs for human-like natural language interfaces"}},"_tags":["story","author_patelajay285","story_15047929","launch_hn"],"author":"patelajay285","children":[15047998,15048291,15048364,15048462,15048464,15048562,15048719,15048995,15050362,15050518,15050539,15050942,15055517],"created_at":"2017-08-18T17:08:54Z","created_at_i":1503076134,"num_comments":47,"objectID":"15047929","points":93,"story_id":15047929,"story_text":"Hi, this is Ajay and Alex, and we\u2019re the founders of Plasticity (https://www.plasticity.ai/). We're building an API that helps developers create human-like natural language interfaces.
Four years ago, we hacked 3rd party commands into Siri without jailbreaking before Alexa Skills or SiriKit were released (https://www.wired.com/2014/04/googolplex/). It was the first App Store for voice commands. Since then, we\u2019ve worked on NL interfaces at Google and Apple Siri. Now we're tackling the next problem: products using NLP are fairly simplistic in what they can do for users. For example, systems like Siri still struggle to directly answer a basic question like "When is the Y Combinator application due?" because it can't understand and reason where the answer may lie in a sentence on Y Combinator's website.
We\u2019re approaching the problem differently by understanding the structure of language and relationships within text, instead of relying on more simplistic methods like keyword matching. We build a graph of entities and their relationships within a sentence along with other linguistic information. You can think of it as \u201cOpen Information Extraction\u201d with a lot more information (https://www.plasticity.ai/api/demo).
Currently, we use a TensorFlow model to perform classical tasks like parts of speech, tokenization, and syntax dependency trees. We built our own Wikipedia crawler for data to better handle chunking and disambiguation, which helps return more accurate results for multi-word entities in sentences like: "The band played let it be by the beatles." We wrote our open IE algorithms from scratch, focusing on speed. It's written completely in C++ and we are adding more features everyday.
Our public APIs are in beta right now, we\u2019re constantly working to improve the accuracy, and we\u2019re looking forward to hearing feedback. We\u2019d love to hear what the HN community is working on with NLP and how we can help!","title":"Launch HN: Plasticity (YC S17) \u2013 APIs for human-like natural language interfaces","updated_at":"2024-09-20T01:14:24Z"}],"hitsPerPage":50,"nbHits":2600,"nbPages":20,"page":0,"params":"query=Andreason&tags=story&hitsPerPage=50&advancedSyntax=true&analyticsTags=backend","processingTimeMS":30,"processingTimingsMS":{"_request":{"roundTrip":19},"afterFetch":{"format":{"highlighting":1,"total":2},"merge":{"mergeLoop":{"prepareNextHit":1,"total":1},"total":1},"total":2},"fetch":{"query":6,"scanning":21,"total":28},"total":30},"query":"Andreason","serverTimeMS":33}