To solve this, we built Reflexes: semantic signals from agent traces, served fast and cheap over API. Built on custom kernels and a custom inference engine forked from vLLM.
Under the hood, it is a small LLM architected around multi-head inference. Small models need to be trained for specific tasks, but running 50 separate small models on the same input for 50 tasks makes no sense.
How it works:\nWe use a modern LLM with hybrid attention and remove the decode step. We built an inference engine that lets prefill compute be 99% reused from reflex to reflex, similar in spirit to older 2019-era BERT/HYDRA and older multiple-head techniques. we built the inference engine to reuse the KV/cache across inputs and compute across all reflexes. One shared backbone reads the trace once, then many heads classify different signals. Our inference engine reuses the same KV/cache and compute across all reflexes, giving us sub-30ms inference with less than 0.1% overhead for each additional reflex.
We took the same high-level idea and did the hard work to make it work with a modern architecture and attention. On it, we can run inference in under 30ms and serve the full request in under 90ms. If you run 4 reflexes or 100, the extra overhead is less than 2ms.
Why does optimizing this matter?
If you\u2019re even a medium-sized startup, you\u2019re dealing with tens of thousands of agent runs and millions of turns. If you want to track things like user frustration rates over time, frontier LLM-as-judge does not scale.
I built a similar stack at Tesla. When ML engineers needed to sample data across petabytes for signals like `is_camera_obfuscated=true`, along with 200 other things, you need to 1) spin them up quickly 2) run at scale efficiently
What it is not:\nA dashboard. 99% of dashboards go unused. \n100% API first and made for devs who want to use this to trigger their own stuff.
vibetrain a custom reflex in our dashboard, and/or then let it self improve in production: https://www.morphllm.com/dashboard/reflex
Docs: https://docs.morphllm.com/sdk/components/reflexes/index
I\u2019d love feedback from people running agents in prod: what sorts of things do you wish you could track over time across 100% of turns but cant right now?
TLDR: semantic signals from agent traces, super fast, cheap via API"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Show HN: Morph Reflexes \u2013 Multi-head classifiers for agent traces"}},"_tags":["story","author_bhaktatejas922","story_48739038","show_hn"],"author":"bhaktatejas922","children":[48739880,48741618,48743196],"created_at":"2026-06-30T20:52:04Z","created_at_i":1782852724,"num_comments":2,"objectID":"48739038","points":20,"story_id":48739038,"story_text":"The most common failures for production agents are behavioral: looping, reasoning leakage, user frustration, and more. Using a frontier model like GPT or Sonnet to judge every turn is too expensive and slow to run at scale.
To solve this, we built Reflexes: semantic signals from agent traces, served fast and cheap over API. Built on custom kernels and a custom inference engine forked from vLLM.
Under the hood, it is a small LLM architected around multi-head inference. Small models need to be trained for specific tasks, but running 50 separate small models on the same input for 50 tasks makes no sense.
How it works:\nWe use a modern LLM with hybrid attention and remove the decode step. We built an inference engine that lets prefill compute be 99% reused from reflex to reflex, similar in spirit to older 2019-era BERT/HYDRA and older multiple-head techniques. we built the inference engine to reuse the KV/cache across inputs and compute across all reflexes. One shared backbone reads the trace once, then many heads classify different signals. Our inference engine reuses the same KV/cache and compute across all reflexes, giving us sub-30ms inference with less than 0.1% overhead for each additional reflex.
We took the same high-level idea and did the hard work to make it work with a modern architecture and attention. On it, we can run inference in under 30ms and serve the full request in under 90ms. If you run 4 reflexes or 100, the extra overhead is less than 2ms.
Why does optimizing this matter?
If you\u2019re even a medium-sized startup, you\u2019re dealing with tens of thousands of agent runs and millions of turns. If you want to track things like user frustration rates over time, frontier LLM-as-judge does not scale.
I built a similar stack at Tesla. When ML engineers needed to sample data across petabytes for signals like `is_camera_obfuscated=true`, along with 200 other things, you need to 1) spin them up quickly 2) run at scale efficiently
What it is not:\nA dashboard. 99% of dashboards go unused. \n100% API first and made for devs who want to use this to trigger their own stuff.
vibetrain a custom reflex in our dashboard, and/or then let it self improve in production: https://www.morphllm.com/dashboard/reflex
Docs: https://docs.morphllm.com/sdk/components/reflexes/index
I\u2019d love feedback from people running agents in prod: what sorts of things do you wish you could track over time across 100% of turns but cant right now?
TLDR: semantic signals from agent traces, super fast, cheap via API","title":"Show HN: Morph Reflexes \u2013 Multi-head classifiers for agent traces","updated_at":"2026-07-09T20:14:16Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"davidbarker"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"I can\u2019t stop watching the \u2018hanger reflex\u2019 freak people out"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://www.inputmag.com/culture/hanger-reflex-makes-head-turn-tiktok-social-media"}},"_tags":["story","author_davidbarker","story_31456389"],"author":"davidbarker","children":[31466324],"created_at":"2022-05-21T09:03:24Z","created_at_i":1653123804,"num_comments":1,"objectID":"31456389","points":20,"story_id":31456389,"title":"I can\u2019t stop watching the \u2018hanger reflex\u2019 freak people out","updated_at":"2024-09-20T11:11:00Z","url":"https://www.inputmag.com/culture/hanger-reflex-makes-head-turn-tiktok-social-media"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"mightybyte"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflex: Practical Functional Reactive Programming (part 1)"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.youtube.com/watch?v=mYvkcskJbc4"}},"_tags":["story","author_mightybyte","story_9334766"],"author":"mightybyte","children":[9334797,9335052],"created_at":"2015-04-07T15:34:53Z","created_at_i":1428420893,"num_comments":2,"objectID":"9334766","points":17,"story_id":9334766,"title":"Reflex: Practical Functional Reactive Programming (part 1)","updated_at":"2023-09-07T02:28:15Z","url":"https://www.youtube.com/watch?v=mYvkcskJbc4"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"TooSimple"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflexio \u2013 An addicting webgame written purely in HTML5 and JavaScript"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://github.com/ketanhwr/reflexio"}},"_tags":["story","author_TooSimple","story_10989368"],"author":"TooSimple","created_at":"2016-01-28T16:58:29Z","created_at_i":1454000309,"num_comments":0,"objectID":"10989368","points":16,"story_id":10989368,"title":"Reflexio \u2013 An addicting webgame written purely in HTML5 and JavaScript","updated_at":"2024-09-19T22:49:39Z","url":"https://github.com/ketanhwr/reflexio"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"heathermiller"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflex: a new programming language for reactive programming by Facebook"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.youtube.com/watch?v=AGkSHE15BSs"}},"_tags":["story","author_heathermiller","story_14745167"],"author":"heathermiller","children":[14745289],"created_at":"2017-07-11T15:51:41Z","created_at_i":1499788301,"num_comments":2,"objectID":"14745167","points":13,"story_id":14745167,"title":"Reflex: a new programming language for reactive programming by Facebook","updated_at":"2024-09-20T01:03:31Z","url":"https://www.youtube.com/watch?v=AGkSHE15BSs"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"DiabloD3"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflex: the return of the high speed skill based FPS"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://www.kickstarter.com/projects/191095869/reflex"}},"_tags":["story","author_DiabloD3","story_8399692"],"author":"DiabloD3","created_at":"2014-10-02T07:10:04Z","created_at_i":1412233804,"num_comments":0,"objectID":"8399692","points":11,"story_id":8399692,"story_text":"","title":"Reflex: the return of the high speed skill based FPS","updated_at":"2023-09-06T22:30:43Z","url":"https://www.kickstarter.com/projects/191095869/reflex"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"jcr"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Human-Machine Interface Keeps Robot Balanced with Human Reflexes"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"http://news.mit.edu/2015/bipedal-robot-with-human-reflexes-0811"}},"_tags":["story","author_jcr","story_10133600"],"author":"jcr","created_at":"2015-08-28T03:29:48Z","created_at_i":1440732588,"num_comments":0,"objectID":"10133600","points":9,"story_id":10133600,"story_text":"","title":"Human-Machine Interface Keeps Robot Balanced with Human Reflexes","updated_at":"2023-09-06T23:29:22Z","url":"http://news.mit.edu/2015/bipedal-robot-with-human-reflexes-0811"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"tim_sw"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflecting on Reflexion: GPT4 AI Agents that self-reflect perform better"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://nanothoughts.substack.com/p/reflecting-on-reflexion"}},"_tags":["story","author_tim_sw","story_35403278"],"author":"tim_sw","children":[35418926,35418928,35451499],"created_at":"2023-04-01T19:27:45Z","created_at_i":1680377265,"num_comments":2,"objectID":"35403278","points":8,"story_id":35403278,"title":"Reflecting on Reflexion: GPT4 AI Agents that self-reflect perform better","updated_at":"2024-09-20T13:43:10Z","url":"https://nanothoughts.substack.com/p/reflecting-on-reflexion"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"fixel"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflex \u2013 New Reactive Programming Language from Facebook"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.youtube.com/watch?v=AGkSHE15BSs"}},"_tags":["story","author_fixel","story_14611436"],"author":"fixel","created_at":"2017-06-22T11:29:19Z","created_at_i":1498130959,"num_comments":0,"objectID":"14611436","points":8,"story_id":14611436,"title":"Reflex \u2013 New Reactive Programming Language from Facebook","updated_at":"2024-09-20T00:55:49Z","url":"https://www.youtube.com/watch?v=AGkSHE15BSs"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"well_i_never"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Dad Reflexes [video]"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://twitter.com/SteveStuWill/status/1140213022259998720"}},"_tags":["story","author_well_i_never","story_20197227"],"author":"well_i_never","children":[20197354],"created_at":"2019-06-16T18:59:19Z","created_at_i":1560711559,"num_comments":2,"objectID":"20197227","points":7,"story_id":20197227,"title":"Dad Reflexes [video]","updated_at":"2024-09-20T04:30:36Z","url":"https://twitter.com/SteveStuWill/status/1140213022259998720"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"apetuskey"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflex v0.8.0 \u2013 Refactoring for Performance"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://reflex.dev/blog/2025-07-01-reflex-080/"}},"_tags":["story","author_apetuskey","story_44638913"],"author":"apetuskey","created_at":"2025-07-21T18:49:39Z","created_at_i":1753123779,"num_comments":0,"objectID":"44638913","points":7,"story_id":44638913,"title":"Reflex v0.8.0 \u2013 Refactoring for Performance","updated_at":"2025-07-22T05:46:16Z","url":"https://reflex.dev/blog/2025-07-01-reflex-080/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"dvt"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflexivity (Social Theory)"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://en.wikipedia.org/wiki/Reflexivity_(social_theory)"}},"_tags":["story","author_dvt","story_35692920"],"author":"dvt","created_at":"2023-04-24T20:39:45Z","created_at_i":1682368785,"num_comments":0,"objectID":"35692920","points":7,"story_id":35692920,"title":"Reflexivity (Social Theory)","updated_at":"2024-09-20T13:54:55Z","url":"https://en.wikipedia.org/wiki/Reflexivity_(social_theory)"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"amichail"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"This Illusion, New to Science, Is Strong Enough to Trick Our Reflexes"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://neurosciencenews.com/expanding-hole-illusion-20710/"}},"_tags":["story","author_amichail","story_31589078"],"author":"amichail","children":[31590696],"created_at":"2022-06-01T22:42:31Z","created_at_i":1654123351,"num_comments":1,"objectID":"31589078","points":6,"story_id":31589078,"title":"This Illusion, New to Science, Is Strong Enough to Trick Our Reflexes","updated_at":"2024-09-20T11:19:19Z","url":"https://neurosciencenews.com/expanding-hole-illusion-20710/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"acidbzeus"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Show HN: Reflexio"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://github.com/ketanhwr/reflexio"}},"_tags":["story","author_acidbzeus","story_12503573","show_hn"],"author":"acidbzeus","children":[12505583],"created_at":"2016-09-15T05:06:18Z","created_at_i":1473915978,"num_comments":1,"objectID":"12503573","points":6,"story_id":12503573,"title":"Show HN: Reflexio","updated_at":"2024-09-19T23:39:33Z","url":"https://github.com/ketanhwr/reflexio"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"schacon"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Why is Git's autocorrect faster than an F1 driver's reflexes?"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://blog.gitbutler.com/why-is-git-autocorrect-too-fast-for-formula-one-drivers/"}},"_tags":["story","author_schacon","story_42686237"],"author":"schacon","created_at":"2025-01-13T17:49:49Z","created_at_i":1736790589,"num_comments":0,"objectID":"42686237","points":6,"story_id":42686237,"title":"Why is Git's autocorrect faster than an F1 driver's reflexes?","updated_at":"2025-01-19T20:38:44Z","url":"https://blog.gitbutler.com/why-is-git-autocorrect-too-fast-for-formula-one-drivers/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"shreyarajpal"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Full stack web apps in Python with Reflex dev"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://techcrunch.com/2023/08/02/reflex-open-source-tool-helps-turn-python-code-into-web-apps/"}},"_tags":["story","author_shreyarajpal","story_36972503"],"author":"shreyarajpal","created_at":"2023-08-02T15:46:49Z","created_at_i":1690991209,"num_comments":0,"objectID":"36972503","points":6,"story_id":36972503,"title":"Full stack web apps in Python with Reflex dev","updated_at":"2024-09-20T14:50:14Z","url":"https://techcrunch.com/2023/08/02/reflex-open-source-tool-helps-turn-python-code-into-web-apps/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"shreyansh2006"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Show HN: Trace the reflexive tool for knowledge workers"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://theprismlabs.github.io/trace/"}},"_tags":["story","author_shreyansh2006","story_44378004","show_hn"],"author":"shreyansh2006","created_at":"2025-06-25T14:49:44Z","created_at_i":1750862984,"num_comments":0,"objectID":"44378004","points":5,"story_id":44378004,"title":"Show HN: Trace the reflexive tool for knowledge workers","updated_at":"2025-06-25T15:55:10Z","url":"https://theprismlabs.github.io/trace/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"pella"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Semmelweis Reflex"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://en.wikipedia.org/wiki/Semmelweis_reflex"}},"_tags":["story","author_pella","story_38170290"],"author":"pella","created_at":"2023-11-06T22:50:43Z","created_at_i":1699311043,"num_comments":0,"objectID":"38170290","points":5,"story_id":38170290,"title":"Semmelweis Reflex","updated_at":"2024-09-20T15:39:17Z","url":"https://en.wikipedia.org/wiki/Semmelweis_reflex"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"guilevi"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflexions of a wannabe software developer"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://guilevi.wordpress.com/2018/04/20/reflexions-of-a-wannabe-software-developer/"}},"_tags":["story","author_guilevi","story_16895951"],"author":"guilevi","created_at":"2018-04-22T12:53:39Z","created_at_i":1524401619,"num_comments":0,"objectID":"16895951","points":5,"story_id":16895951,"title":"Reflexions of a wannabe software developer","updated_at":"2024-09-20T02:25:17Z","url":"https://guilevi.wordpress.com/2018/04/20/reflexions-of-a-wannabe-software-developer/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"kgoutham93"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflex: Reactive Programming at Facebook [video]"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.youtube.com/watch?v=AGkSHE15BSs"}},"_tags":["story","author_kgoutham93","story_14619740"],"author":"kgoutham93","created_at":"2017-06-23T15:15:39Z","created_at_i":1498230939,"num_comments":0,"objectID":"14619740","points":5,"story_id":14619740,"title":"Reflex: Reactive Programming at Facebook [video]","updated_at":"2024-09-20T00:56:37Z","url":"https://www.youtube.com/watch?v=AGkSHE15BSs"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"drm237"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Social Graph visibility akin to pain reflex"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"http://radar.oreilly.com/archives/2008/02/social_graph_pain_reflex.html"}},"_tags":["story","author_drm237","story_108474"],"author":"drm237","created_at":"2008-02-02T22:28:51Z","created_at_i":1201991331,"num_comments":0,"objectID":"108474","points":5,"story_id":108474,"title":"Social Graph visibility akin to pain reflex","updated_at":"2023-09-06T19:08:31Z","url":"http://radar.oreilly.com/archives/2008/02/social_graph_pain_reflex.html"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"ypavan"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Semmelweis reflex"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"http://en.wikipedia.org/wiki/Semmelweis_reflex"}},"_tags":["story","author_ypavan","story_926511"],"author":"ypavan","children":[926513],"created_at":"2009-11-06T18:01:55Z","created_at_i":1257530515,"num_comments":2,"objectID":"926511","points":4,"story_id":926511,"story_text":"","title":"Semmelweis reflex","updated_at":"2024-09-19T16:47:29Z","url":"http://en.wikipedia.org/wiki/Semmelweis_reflex"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"mightybyte"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflex: Practical FRP (part 2)"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.youtube.com/watch?v=3qfc9XFVo2c"}},"_tags":["story","author_mightybyte","story_9335828"],"author":"mightybyte","children":[9335850],"created_at":"2015-04-07T18:00:10Z","created_at_i":1428429610,"num_comments":1,"objectID":"9335828","points":4,"story_id":9335828,"story_text":"","title":"Reflex: Practical FRP (part 2)","updated_at":"2023-09-07T02:28:20Z","url":"https://www.youtube.com/watch?v=3qfc9XFVo2c"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"aiNohY6g"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflex: Web Apps in Pure Python"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://reflex.dev/"}},"_tags":["story","author_aiNohY6g","story_37952714"],"author":"aiNohY6g","children":[37952888],"created_at":"2023-10-20T05:47:41Z","created_at_i":1697780861,"num_comments":1,"objectID":"37952714","points":4,"story_id":37952714,"title":"Reflex: Web Apps in Pure Python","updated_at":"2024-09-20T15:26:04Z","url":"https://reflex.dev/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"numlocked"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflexion: An autonomous agent with dynamic memory and self-reflection"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://arxiv.org/abs/2303.11366"}},"_tags":["story","author_numlocked","story_35314773"],"author":"numlocked","children":[35317932],"created_at":"2023-03-26T14:27:04Z","created_at_i":1679840824,"num_comments":1,"objectID":"35314773","points":4,"story_id":35314773,"title":"Reflexion: An autonomous agent with dynamic memory and self-reflection","updated_at":"2024-09-20T13:34:55Z","url":"https://arxiv.org/abs/2303.11366"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"mustafa01ali"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflex: An architecture for building mobile apps with agents"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://shopify.engineering/reflex"}},"_tags":["story","author_mustafa01ali","story_50380949"],"author":"mustafa01ali","created_at":"2026-11-05T16:14:12Z","created_at_i":1793895252,"num_comments":0,"objectID":"50380949","points":4,"story_id":50380949,"title":"Reflex: An architecture for building mobile apps with agents","updated_at":"2026-11-10T03:36:38Z","url":"https://shopify.engineering/reflex"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"birdculture"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"My reflex of assuming our servers need ever-growing amounts of RAM"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://utcc.utoronto.ca/~cks/space/blog/sysadmin/ServerRAMGrowthReflex"}},"_tags":["story","author_birdculture","story_50346276"],"author":"birdculture","created_at":"2026-11-03T00:23:54Z","created_at_i":1793665434,"num_comments":0,"objectID":"50346276","points":4,"story_id":50346276,"title":"My reflex of assuming our servers need ever-growing amounts of RAM","updated_at":"2026-11-03T04:03:14Z","url":"https://utcc.utoronto.ca/~cks/space/blog/sysadmin/ServerRAMGrowthReflex"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"vismit2000"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Photic Sneeze Reflex"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://en.wikipedia.org/wiki/Photic_sneeze_reflex"}},"_tags":["story","author_vismit2000","story_49669975"],"author":"vismit2000","created_at":"2026-09-12T07:45:27Z","created_at_i":1789199127,"num_comments":0,"objectID":"49669975","points":4,"story_id":49669975,"title":"Photic Sneeze Reflex","updated_at":"2026-09-24T04:02:51Z","url":"https://en.wikipedia.org/wiki/Photic_sneeze_reflex"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"danw"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Where's Chris? (on reflexive design): \"If I won't use it, I shouldn't expect you to either\""},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"http://www.purselipsquarejaw.org/2004/10/wheres-chris-on-reflexive-design.php"}},"_tags":["story","author_danw","story_48887"],"author":"danw","created_at":"2007-08-31T19:23:18Z","created_at_i":1188588198,"num_comments":0,"objectID":"48887","points":4,"story_id":48887,"title":"Where's Chris? (on reflexive design): \"If I won't use it, I shouldn't expect you to either\"","updated_at":"2024-09-19T16:21:31Z","url":"http://www.purselipsquarejaw.org/2004/10/wheres-chris-on-reflexive-design.php"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"memorable"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"LatencyFleX: An Alternative to Nvidia Reflex"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/ishitatsuyuki/LatencyFleX"}},"_tags":["story","author_memorable","story_35665438"],"author":"memorable","created_at":"2023-04-22T11:29:40Z","created_at_i":1682162980,"num_comments":0,"objectID":"35665438","points":4,"story_id":35665438,"title":"LatencyFleX: An Alternative to Nvidia Reflex","updated_at":"2024-09-20T13:52:12Z","url":"https://github.com/ishitatsuyuki/LatencyFleX"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"nrvous"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Self-Reflexive Instance-Naming"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://blog.saxifrage.one/recursion/linguistics/egcoin/"}},"_tags":["story","author_nrvous","story_33846407"],"author":"nrvous","created_at":"2022-12-03T19:23:14Z","created_at_i":1670095394,"num_comments":0,"objectID":"33846407","points":4,"story_id":33846407,"title":"Self-Reflexive Instance-Naming","updated_at":"2024-09-20T12:39:33Z","url":"https://blog.saxifrage.one/recursion/linguistics/egcoin/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"jger15"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflexivity in Tech Returns: Talent Chasing Money Chasing Talent"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://www.thediff.co/p/reflexivity-in-tech-returns-talent"}},"_tags":["story","author_jger15","story_30024561"],"author":"jger15","created_at":"2022-01-21T15:13:54Z","created_at_i":1642778034,"num_comments":0,"objectID":"30024561","points":4,"story_id":30024561,"title":"Reflexivity in Tech Returns: Talent Chasing Money Chasing Talent","updated_at":"2024-09-20T10:17:24Z","url":"https://www.thediff.co/p/reflexivity-in-tech-returns-talent"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"sytse"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"The Origin of Stimulus Reflex"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://stimulusjs.org/handbook/origin"}},"_tags":["story","author_sytse","story_22799240"],"author":"sytse","created_at":"2020-04-07T00:21:47Z","created_at_i":1586218907,"num_comments":0,"objectID":"22799240","points":4,"story_id":22799240,"title":"The Origin of Stimulus Reflex","updated_at":"2024-09-20T05:58:44Z","url":"https://stimulusjs.org/handbook/origin"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"milgra"},"story_text":{"matchLevel":"none","matchedWords":[],"value":"Converted one of my C/OpenGL games to WebAssembly/WebGL with Emscripten, it was surprisingly easy and seamless."},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Show HN: Cortex \u2013 WebAssembly/WebGL reflex game"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://milgra.com/cortex/"}},"_tags":["story","author_milgra","story_35337861","show_hn"],"author":"milgra","children":[35338141],"created_at":"2023-03-28T09:22:38Z","created_at_i":1679995358,"num_comments":3,"objectID":"35337861","points":3,"story_id":35337861,"story_text":"Converted one of my C/OpenGL games to WebAssembly/WebGL with Emscripten, it was surprisingly easy and seamless.","title":"Show HN: Cortex \u2013 WebAssembly/WebGL reflex game","updated_at":"2024-09-20T13:37:27Z","url":"https://milgra.com/cortex/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Bluestein"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"The Minsky Moment and Theory of Reflexivity: Minsky and Soros About Instability"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://speculatorsanonymous.com/articles/minsky-moment-theory-of-reflexivity/"}},"_tags":["story","author_Bluestein","story_40820023"],"author":"Bluestein","children":[40820500],"created_at":"2024-06-28T12:36:08Z","created_at_i":1719578168,"num_comments":2,"objectID":"40820023","points":3,"story_id":40820023,"title":"The Minsky Moment and Theory of Reflexivity: Minsky and Soros About Instability","updated_at":"2024-09-20T17:20:27Z","url":"https://speculatorsanonymous.com/articles/minsky-moment-theory-of-reflexivity/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"romanfll"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflexive Memory: A CPU-Only Alternative to Transformers for Streaming Data"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://medium.com/towards-artificial-intelligence/reflexive-memory-a-cpu-only-alternative-to-transformers-for-streaming-forecasting-45efc12e383c"}},"_tags":["story","author_romanfll","story_46543300"],"author":"romanfll","children":[46543301],"created_at":"2026-01-08T16:54:27Z","created_at_i":1767891267,"num_comments":1,"objectID":"46543300","points":3,"story_id":46543300,"title":"Reflexive Memory: A CPU-Only Alternative to Transformers for Streaming Data","updated_at":"2026-03-05T23:19:21Z","url":"https://medium.com/towards-artificial-intelligence/reflexive-memory-a-cpu-only-alternative-to-transformers-for-streaming-forecasting-45efc12e383c"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"sabi_soltani"},"story_text":{"matchLevel":"none","matchedWords":[],"value":"We built Golden Owl because the data that matters to strategy, like leaks, emerging competitors, reputation hits, and regulatory shifts, is scattered and hard to catch in time.
Golden Owl uses a network of AI agents to monitor and analyze data across the open, deep, and dark web. It pulls from non-indexed sources like court filings, breach dumps, sanctions lists, forum threads, and social media. Then it de-duplicates, merges, and ranks everything in real time.
What it does:
Discover: Finds hidden documents, domains, leaks, forum posts, and more from hundreds of OSINT sources.
Analyze: Uses AI to extract entities, map relationships, detect anomalies, and identify trends.
Monitor: Sends live alerts on brand mentions, competitor actions, geopolitical shifts, and supply chain risks.
Compare: Provides side-by-side dashboards for companies, markets, or investments.
Under the hood:
~200 AI agents act like digital investigators at scale.
Custom scrapers + Tor relays for deep/dark web access.
Vector graph store for relationships and clustering.
API with 100+ endpoints offering both raw and enriched data.
All data stays private\u2014no public indexes, no shared queries.
We're live at https://goldenowl.ai
Would love feedback on:\nUsability and UI\nData coverage and freshness\nAnything you'd expect from an OSINT tool but don't see
Thanks!"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Show HN: Reflexive AI Intelligence hub replicating deep human investigations"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://goldenowl.ai"}},"_tags":["story","author_sabi_soltani","story_44044117","show_hn"],"author":"sabi_soltani","children":[44155459],"created_at":"2025-05-20T17:54:36Z","created_at_i":1747763676,"num_comments":1,"objectID":"44044117","points":3,"story_id":44044117,"story_text":"We built Golden Owl because the data that matters to strategy, like leaks, emerging competitors, reputation hits, and regulatory shifts, is scattered and hard to catch in time.
Golden Owl uses a network of AI agents to monitor and analyze data across the open, deep, and dark web. It pulls from non-indexed sources like court filings, breach dumps, sanctions lists, forum threads, and social media. Then it de-duplicates, merges, and ranks everything in real time.
What it does:
Discover: Finds hidden documents, domains, leaks, forum posts, and more from hundreds of OSINT sources.
Analyze: Uses AI to extract entities, map relationships, detect anomalies, and identify trends.
Monitor: Sends live alerts on brand mentions, competitor actions, geopolitical shifts, and supply chain risks.
Compare: Provides side-by-side dashboards for companies, markets, or investments.
Under the hood:
~200 AI agents act like digital investigators at scale.
Custom scrapers + Tor relays for deep/dark web access.
Vector graph store for relationships and clustering.
API with 100+ endpoints offering both raw and enriched data.
All data stays private\u2014no public indexes, no shared queries.
We're live at https://goldenowl.ai
Would love feedback on:\nUsability and UI\nData coverage and freshness\nAnything you'd expect from an OSINT tool but don't see
Thanks!","title":"Show HN: Reflexive AI Intelligence hub replicating deep human investigations","updated_at":"2025-12-13T21:39:10Z","url":"https://goldenowl.ai"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"yamrzou"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Semmelweis Reflex"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://en.wikipedia.org/wiki/Semmelweis_reflex"}},"_tags":["story","author_yamrzou","story_36323923"],"author":"yamrzou","children":[36324021],"created_at":"2023-06-14T10:43:38Z","created_at_i":1686739418,"num_comments":1,"objectID":"36323923","points":3,"story_id":36323923,"title":"Semmelweis Reflex","updated_at":"2024-09-20T14:21:27Z","url":"https://en.wikipedia.org/wiki/Semmelweis_reflex"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"sharemywin"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Types of AI Agents \u2013 Reflexive, Model, Goal, Utility, and Learning"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://tutorialforbeginner.com/types-of-artificial-intelligence"}},"_tags":["story","author_sharemywin","story_35631443"],"author":"sharemywin","children":[35631451],"created_at":"2023-04-19T18:16:56Z","created_at_i":1681928216,"num_comments":1,"objectID":"35631443","points":3,"story_id":35631443,"title":"Types of AI Agents \u2013 Reflexive, Model, Goal, Utility, and Learning","updated_at":"2024-09-20T13:48:36Z","url":"https://tutorialforbeginner.com/types-of-artificial-intelligence"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"pndy"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Photic Sneeze Reflex"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://en.wikipedia.org/wiki/Photic_sneeze_reflex"}},"_tags":["story","author_pndy","story_35450343"],"author":"pndy","children":[35450365],"created_at":"2023-04-05T06:57:07Z","created_at_i":1680677827,"num_comments":1,"objectID":"35450343","points":3,"story_id":35450343,"title":"Photic Sneeze Reflex","updated_at":"2024-09-20T13:48:38Z","url":"https://en.wikipedia.org/wiki/Photic_sneeze_reflex"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"rockstar9"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflexive is acquired by Amazon.com"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"http://www.reflexive.com/index.php?PAGE=Blog&BID=169"}},"_tags":["story","author_rockstar9","story_339922"],"author":"rockstar9","children":[339952],"created_at":"2008-10-22T11:30:42Z","created_at_i":1224675042,"num_comments":1,"objectID":"339922","points":3,"story_id":339922,"title":"Reflexive is acquired by Amazon.com","updated_at":"2023-09-06T19:26:07Z","url":"http://www.reflexive.com/index.php?PAGE=Blog&BID=169"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"tosh"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"Reflexive McLuhanism"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["reflex"],"value":"https://www.swyx.io/reflexive-mcluhanism/"}},"_tags":["story","author_tosh","story_26764050"],"author":"tosh","children":[26764054,26764876],"created_at":"2021-04-10T19:00:14Z","created_at_i":1618081214,"num_comments":1,"objectID":"26764050","points":3,"story_id":26764050,"title":"Reflexive McLuhanism","updated_at":"2024-09-20T08:22:17Z","url":"https://www.swyx.io/reflexive-mcluhanism/"}],"hitsPerPage":50,"nbHits":6785,"nbPages":20,"page":0,"params":"query=Reflex&tags=story&hitsPerPage=50&advancedSyntax=true&analyticsTags=backend","processingTimeMS":11,"processingTimingsMS":{"_request":{"roundTrip":21},"afterFetch":{"format":{"total":1},"merge":{"mergeLoop":{"prepareNextHit":1,"total":2},"total":2},"total":2},"fetch":{"query":3,"scanning":4,"total":8},"total":11},"query":"Reflex","serverTimeMS":13}