@rsdgpt Avatar @rsdgpt Ryan D’Onofrio

Ryan D’Onofrio posts on X about agent, ai, build, open ai the most. They currently have [---] followers and [---] posts still getting attention that total [------] engagements in the last [--] hours.

Engagements: [------] #

Engagements Line Chart

Mentions: [--] #

Mentions Line Chart

Followers: [---] #

Followers Line Chart

CreatorRank: undefined #

CreatorRank Line Chart

Social Influence

Social category influence technology brands stocks finance social networks vc firms celebrities

Social topic influence agent, ai, build, open ai, api, if you, future, tools for, super bowl, 6969

Top accounts mentioned or mentioned by @benhylak @raindropai @rtcgpt @thdxr @ryandotelliott @aaxsh18 @mixedbreadai @bcmerchant @openai @theo @pallavmac @maziyarpanahi @antoinechaffin @din0s @verceldev @aidanmclau @vercel @therealadamg @tekbog @jeremyphoward

Top assets mentioned Frontier (FRONT) Microsoft Corp. (MSFT) Spotify Technology (SPOT) Alphabet Inc Class A (GOOGL) Reddit, Inc. (RDDT) Ontology (ONT) Opus (OPUS)

Top Social Posts

Top posts by engagements in the last [--] hours

"@rtcgpt I wonder if we will see a pivot in the modeling and marketing if that is the case. OAI continues to claim AGI for all but if their product lacks integrations into the things we use to run lives/businesses will they pivot or just hope Microsoft wins the battle for our eyes"
X Link 2023-12-08T03:08Z [--] followers, [--] engagements

"OpenAI losing the lead Or calm before the 🌩🌩"
X Link 2023-12-11T19:35Z [--] followers, [--] engagements

"@PicoPaco17 So what do you recommend people be building with AI right now How do new companies get started"
X Link 2023-12-17T17:02Z [--] followers, [--] engagements

"Anybody know of any models fine tuned for generating stable diffusion prompts Ideally for control net but flexible"
X Link 2023-12-30T21:02Z [--] followers, [--] engagements

"Whos building stable diffusion prompt upscalers Feel like so much alpha here for coherence. Why hasnt this happened"
X Link 2024-01-04T01:36Z [--] followers, [--] engagements

"Wanted to do something fun so tonight built a quick little program that from any Spotify link makes Lyrics. I figure a great tool for those who like indie music Used #Spotify #OpenAI #Whisper #buildinpublic"
X Link 2024-01-23T05:14Z [--] followers, [--] engagements

"@alonsosilva Wow. This is impressive. And incredibly easy to get other tokens values I imagine as well right Feed it a few examples you know the answer to and save those for future use. Think so much value here for predictable outcomes and especially function selection"
X Link 2024-01-24T18:11Z [--] followers, [--] engagements

"Fine. Typescript and React are pretty incredible especially for devs like myself who cant store everything in their head well. It literally tells you whats busted and you go clean it"
X Link 2024-02-23T17:31Z [--] followers, [--] engagements

"Hey @fedjabosnic working on a project as part of a Venture Capitalist competition at my university and would love to chat This could be huge for us Anyone starting a new saas project any time soon Looking for a startup that we can build out a blog for: we will boost your organic traffic and Google rankings and grow your DR - even before your product is ready. You can be hands off and let us do it. Anyone starting a new saas project any time soon Looking for a startup that we can build out a blog for: we will boost your organic traffic and Google rankings and grow your DR - even before your"
X Link 2024-03-02T19:44Z [--] followers, [--] engagements

"This is the third thing in AI to make my mouth fall open. GPT-4 Generative Fill now Devin. Congrats to the team. Cant wait to try. 4/4 We even tried giving Devin real jobs on Upwork and it could do those too https://t.co/7XUWQLeZm1 4/4 We even tried giving Devin real jobs on Upwork and it could do those too https://t.co/7XUWQLeZm1"
X Link 2024-03-12T16:00Z [--] followers, [---] engagements

"Need Yann Altman and Hinton to sit in a room and develop an operational definition of AGI. Otherwise OAI will scream AGI when it can beat PhDs at trivia and Yann will say it cant see at the same time so therefore not AGI"
X Link 2024-03-21T22:28Z [--] followers, [--] engagements

"Just submitted my first application for Y Combinator. Dont think my company is quite ready yet but hey you never know"
X Link 2024-04-09T14:33Z [--] followers, [--] engagements

"@mattpocockuk I recently built my first app with Typescript/NextJS. Its not very performative and I think Im doing a lot inefficiently. How do you recommend moving through it and improving Reducing prop drilling/ excessive use state hooks all over etc. Thanks"
X Link 2024-04-28T15:20Z [---] followers, [--] engagements

"@rtcgpt Wait this is sick. Reminds me of the "Bill Clinton" meme but way higher quality. signing up"
X Link 2024-05-01T04:45Z [---] followers, [--] engagements

"GPT-4 for Python. Opus for Typescript"
X Link 2024-05-08T14:57Z [--] followers, [--] engagements

"First hour coding with GPT4o. Its cracked"
X Link 2024-05-13T20:49Z [--] followers, [--] engagements

"Want to believe if they are making this free and available to all NOW that the capabilities of [--] are a huge leap forward on the frontier"
X Link 2024-05-13T21:54Z [--] followers, [--] engagements

"@_GreatDelusion @nxthompson @TheAtlantic @OpenAI No. Spoken live at a UC with Nita Farahany"
X Link 2024-05-30T01:31Z [--] followers, [--] engagements

"Project based learning is the only way. Eureka started as a Flask App then a FastAPI app then a React App and now a NextJS app. Working on one project helped teach me four new libraries and skills. [--] months ago I had never touched TypeScript. Just make cool stuff"
X Link 2024-06-09T00:11Z [---] followers, [---] engagements

"@rtcgpt Okay its a little novel so maybe hop in communities on Reddit and X related. Show off a demo and start a wait-list to measure demand"
X Link 2024-06-11T21:03Z [---] followers, [--] engagements

"Shadcn mobile πŸ‘€πŸ‘€"
X Link 2024-08-30T15:38Z [---] followers, [--] engagements

"As someone who ships with Vite this is huge for me LOL First up the CLI now has support for all major React frameworks out of the box: Next.js Remix Vite and Laravel. And when you initialize a new app we update your existing Tailwind files instead of overriding them. https://t.co/D6cuJV98iV First up the CLI now has support for all major React frameworks out of the box: Next.js Remix Vite and Laravel. And when you initialize a new app we update your existing Tailwind files instead of overriding them. https://t.co/D6cuJV98iV"
X Link 2024-08-30T22:54Z [---] followers, [---] engagements

"Have to assume Sama/Satya will say something on Opera re: next model. It's middle of funding round and they've been silent for months on frontier"
X Link 2024-08-31T18:03Z [---] followers, [--] engagements

"This was me when I deployed four projects on the same CIDR block and a Senior engineer had to help me destroy [--] of them. What a great senior dev πŸ’ͺ🏻 https://t.co/pQDES5srNn What a great senior dev πŸ’ͺ🏻 https://t.co/pQDES5srNn"
X Link 2024-08-31T21:41Z [---] followers, [--] engagements

"OpenAI boutta drop the hottest blog post of [----]. πŸ›³ πŸ›³"
X Link 2024-09-05T04:19Z [---] followers, [---] engagements

"@skalskip92 Bro is the goat. Cant wait to play with this ASAP Im gonna need a bigger GPU"
X Link 2024-09-12T16:20Z [---] followers, [---] engagements

"Kinda crazy that dev day is in [--] days what will OAI show"
X Link 2024-09-20T16:02Z [---] followers, [---] engagements

"Dudes be like we created an ontology and the ontology in question be like"
X Link 2024-11-13T02:05Z [---] followers, [--] engagements

"Late night walks actually the best for working on hard programming problems. Brain can whirl and quickly write little messages to chatGPT for later"
X Link 2024-12-10T05:52Z [--] followers, [--] engagements

"@aidan_mclau But if thats really the next step way smaller change than 3.5- [--] transition. Even if capabilities /benchmarks logarithmic feels like really minor push to me. Like if thats the 10x huge diminishing returns. :/"
X Link 2024-12-17T17:46Z [--] followers, [----] engagements

"@aidan_mclau True. Anthropic being silent in all this. If you had to guess something cooking or side effect of not wanting to release [---] Opus due to the performance/price wrt NewSonnet"
X Link 2024-12-17T17:50Z [--] followers, [----] engagements

"@theo Im confused. How has anything youve done changed this particular metric Isnt this just the public API vs whatever ChatGPT serves The only place I can imagine is time from receiving the partial response to updating UI but even that doesnt seem to explain this gap. Thanks"
X Link 2025-01-13T03:55Z [--] followers, [----] engagements

"@OpenAIDevs in chatgpt about 50% of my o3/o3-m-h requests have been stuck hanging while "reasoning" last [--] hours"
X Link 2025-02-14T19:37Z [--] followers, [--] engagements

"1.5 hours from my wife asking me to build her a word hunt game to deployed PWA on her phone. Planning stage with o1. Created todo list and had Sonnet [---] execute then deployed on @vercel. @shadcn components. Dev cycles getting SO fast. Personalized software is here"
X Link 2025-03-02T22:52Z [--] followers, [----] engagements

"@theo Agreed they raise the ceiling. Everyone else raising the floor atm"
X Link 2025-03-30T22:28Z [--] followers, [----] engagements

"@TheRealAdamG @simonw No release notes. No new names. Not clear whats changed. I would not call that a model release king"
X Link 2025-04-26T01:15Z [---] followers, [--] engagements

"@RyanDotElliott Sequoia Capital is calling. Will you answer the call"
X Link 2025-05-01T14:36Z [---] followers, [--] engagements

"@tekbog I write non-OOP Python AWS LLM slop. Plz hire"
X Link 2025-06-07T03:23Z [---] followers, [--] engagements

"Fresh chat with Sonnet [--] provided [--] relevant files and outlined my bug vs expected behavior (deployment configuration) it went and found the actual issue in my CI which I hadnt even considered. Feeling models start to go BEYOND whats provided in immediate context so magic"
X Link 2025-06-10T16:23Z [---] followers, [--] engagements

"@jeremyphoward @karpathy I LOVE LLMs.txt. that + deepwiki (then porting to markdown) have really changed how I give my LLM context for libraries outside the norm"
X Link 2025-06-19T04:10Z [---] followers, [--] engagements

"@ryandavogel @rauchg @v0 i didnt even realize that it would continue to show the updates as chat progressed. that's awesome. v0 becoming such a great platform for fast iteration and exploring an idea. this one let the ai get too carried away here so gotta roll back"
X Link 2025-06-22T17:24Z [---] followers, [----] engagements

"i dream of a world where i can upload a folder of data explain the type of ML model i want to make and an LLM can go train my model. Tasks like data sorting setting up the pipeline etc abstracted away. like @roboflow meets @raindrop_ai . idk @benhylak is this anything Vercel Sandbox securely runs untrusted code in isolated cloud environments like AI-generated code. Create ephemeral isolated microVMs using the new Sandbox SDK with up to 45m execution times. Now in Beta and available to customers on all plans. https://t.co/QSvhJjayep Vercel Sandbox securely runs untrusted code in isolated cloud"
X Link 2025-06-25T20:25Z [---] followers, [----] engagements

"Deep research API to sports betting pipeline"
X Link 2025-06-26T17:51Z [---] followers, [--] engagements

"New personal site is live. Shows off my favorite games you can actually play Pokmon Blackjack and Chess. and every asset is 100% AI-generated. As always built with Next.js"
X Link 2025-07-05T22:56Z [---] followers, [--] engagements

"The process started with the base image. Fed a bunch of inspo photos to a few models but ended up using GPT-4o to get a workable canvas. Then came the video. MANY prompting iterations with Bytedance's Seedance-1-pro. The model constantly hallucinates breaks physics adds extra hands. You have to brute-force every single animation"
X Link 2025-07-05T22:56Z [---] followers, [--] engagements

"@thdxr 0.1.121"
X Link 2025-07-08T23:33Z [---] followers, [---] engagements

"@pallavmac the little crackle insane"
X Link 2025-07-11T16:10Z [---] followers, [--] engagements

"@tunguz yeah who would build a hot-dog or not app. some sick world we live in. sick i say. https://hotdog-nu-six.vercel.app/ https://hotdog-nu-six.vercel.app/"
X Link 2025-07-11T22:55Z [---] followers, [---] engagements

"@rtcgpt Gemini [--] probably"
X Link 2025-07-13T20:46Z [---] followers, [--] engagements

"@OpenAI how does this reconcile with these papers and Qwen paper on prefilling with wait producing the same results Seems like the labs are saying [--] different things on CoT https://www.anthropic.com/research/reasoning-models-dont-say-think https://www.anthropic.com/research/reasoning-models-dont-say-think"
X Link 2025-07-15T16:24Z [---] followers, [----] engagements

"@Antisimplistic @OpenAI yes i see that. but I would think at some of the most important companies of our time [--] employees might have a professional disagreement on approach in a brand new field"
X Link 2025-07-15T16:33Z [---] followers, [--] engagements

"πŸ‘€πŸ‘€ fast good tokens are STILL underrated right now. Imagine Claude [--] intelligence at these speeds. Software would be on demand Qwen3-Coder at [----] tokens/sec is now live in Windsurf ⚑ Fully hosted on US servers by @CerebrasSystems. Video is 1x speed. https://t.co/PbHuenE0Nu Qwen3-Coder at [----] tokens/sec is now live in Windsurf ⚑ Fully hosted on US servers by @CerebrasSystems. Video is 1x speed. https://t.co/PbHuenE0Nu"
X Link 2025-08-01T18:01Z [---] followers, [--] engagements

"A man goes to a doctor Doctor Im depressed life is harsh unforgiving cruel The doctor lights up. The treatment after all is simple. Its going to be a big week the doctor says new models and features galore That should sort you out. The man bursts into tears. But doctor he says nothing ever happens"
X Link 2025-08-04T17:24Z [---] followers, [--] engagements

"@OfficialLoganK hey Logan my company is looking to move about 200M monthly inference calls off AWS to Gemini. But operating with really strict HIPAA and data retention policies. Any recommendation on best route to go down or someone to connect w/ at Gemini Thanks"
X Link 2025-08-13T18:22Z [---] followers, [---] engagements

"@benhylak impossible. they are dedicated to making the worst llm experience at every turn"
X Link 2025-08-14T22:37Z [---] followers, [--] engagements

"@NateSilver538 Think the advantage of academic journals is still that ideas have to be further fleshed out. You have to GO somewhere in many that rigor isnt enforced other places. Agree with usual points on PR originality quality but need something a little more complete than substack imo"
X Link 2025-08-17T16:21Z [---] followers, [----] engagements

"a small part of me thinks the frontier labs are seeing early signs of self improvement and going to slow down on releases to RL the living hell out of AI development"
X Link 2025-08-20T05:19Z [---] followers, [--] engagements

"@Miles_Brundage In your mind is this hedging bets against a permissive and flip floppy Trump administration since during Bidens tenure OAI via sama was begging to be regulated Or more due to the competition tightening"
X Link 2025-08-25T16:36Z [---] followers, [---] engagements

"@OpenRouterAI feature suggestion: your AI multi chat is great. would be even greater to one click "export to code" so I could have a python script that calls my mutli models (with my settings) and gets the response Vertex AI has one like this and it's really useful"
X Link 2025-09-01T22:27Z [---] followers, [--] engagements

"Put [--] SOTA LLMs to the test: write genetic algorithms to train a 2-layer MLP to play Snake. Same constraints [---] generations let evolution decide"
X Link 2025-09-02T14:39Z [---] followers, [---] engagements

"Results: GPT-5: [----] score [---] lines (wins but overcomplicated) Grok Code: [----] score [--] lines (elegantly simple) DeepSeek: [----] score [--] lines (best efficiency) Mistral: [---] score [---] lines (fancy code zero learning)"
X Link 2025-09-02T14:39Z [---] followers, [---] engagements

"@tekbog Funny thing is lambda end of life for Python [---] is next month right Ah AWS"
X Link 2025-09-05T06:11Z [---] followers, [--] engagements

"Whos making the AWS RL environment"
X Link 2025-09-15T20:15Z [---] followers, [--] engagements

"@MaziyarPanahi @huggingface @OpenMed_AI Interested in running one of these at massive scale. [---] pages a second. 300m pages per month. Possible to chat on architecture"
X Link 2025-09-15T21:50Z [---] followers, [--] engagements

"Initial Orchid impressions. Solid UI's got this in one shot with a pretty short prompt. @kevinlu625 @orchidsapp"
X Link 2025-09-18T02:45Z [---] followers, [--] engagements

"Unfortunately both times I have tried multi-turn chats with Orchid it gets stuck "generating" and never completes. I'm hopeful for the product but I can't get to a completed UI. Would also love more "prompt optimizers" for the great UIs I see it CAN make"
X Link 2025-09-18T02:45Z [---] followers, [--] engagements

"I will never get tired of the rubber ducky in my pocket. I word vomit my thoughts my ideas my constraints code scaffolding in natural language. And instantly do a deep dive on any part of it. Incredible"
X Link 2025-09-22T18:08Z [---] followers, [--] engagements

"@benhylak Agreed. So so sick of it just saying shit. What does any of this mean It gets so distracted explodes random jargon and not actionable plans. Im ready to cancel if it wasnt for codex rate limits rn"
X Link 2025-09-22T18:29Z [---] followers, [----] engagements

"I gave both GPT-5 and Claude [--] access to nano-banana and encouraged them to express themselves. seems about right"
X Link 2025-09-23T22:15Z [---] followers, [---] engagements

"@MaziyarPanahi sorry if i was unclear. im interested just in labeling my [--] categories like whatever NER model can do those broad categories generally is useful to me"
X Link 2025-09-30T21:23Z [---] followers, [--] engagements

"@dan_goosewin Real ones have also dealt with thinking their changes to the UI didnt work but it simply hasnt propagated yet but you freak out for [--] minutes convinced you pushed into the air"
X Link 2025-10-01T17:04Z [---] followers, [----] engagements

"What are yall using for async agents Like web app kicks off task. Agent needs access to files info tools etc. where do you have it do the work and then update your UI when complete"
X Link 2025-10-08T04:23Z [---] followers, [--] engagements

"@RyanDotElliott Parakeet from NVIDIA In the future might allow people to choose from a list"
X Link 2025-10-13T20:33Z [---] followers, [--] engagements

"@reductoai @a16z Wow congratulations Huge testament to great engineering and strong vision for the future Excited to see what is next for the team Rooting for you"
X Link 2025-10-14T16:14Z [---] followers, [---] engagements

"This should have 1k RTs. This is TERRIFYING for your LLM as a judge Evals. Further proof best thing you can do is [--]. Human annotation [--]. Error analysis. We have been testing LLM/VLM-as-judge for our new models. We found that certain models show higher position consistency than others. For example we ran the same test set (n=253) with Grok-4-fast-reasoning and GPT-5-thinking-high. For each model we ran the same prompt twice with https://t.co/ppwQDiUdGJ We have been testing LLM/VLM-as-judge for our new models. We found that certain models show higher position consistency than others. For"
X Link 2025-10-15T03:24Z [---] followers, [--] engagements

"If youre 20-30 and your main circle isnt discussing the new YouTube UI then its time to find a new circle Your network is your net worth"
X Link 2025-10-15T03:50Z [---] followers, [---] engagements

"@_WEEXIAO @Osmosis_AI Congrats team Excited to learn more next week and see how it might enable new capabilities in the healthcare space"
X Link 2025-10-15T17:56Z [---] followers, [---] engagements

"I built osgrep Its a local code search tool that understands natural language. Works as a standalone CLI or a plugin for Claude Code. No API keys or subscription. I wanted the power of "semantic search" without the latency price or privacy trade-offs. Video in realtime"
X Link 2025-11-23T01:25Z [---] followers, 69.9K engagements

"@SullyOmarr I need AWS to have a model like this so bad lol. Building a research assistant on huge pdfs and knowing theres [--] million CW out there hurts my little 200k heart"
X Link 2025-02-07T16:27Z [--] followers, 19K engagements

"@jeremyphoward okay modern-Bert is incredible. Was able to get SOTA in just [--] epochs of training on NER disease task. Question want to also do symptom and personal information recognition. Better to train several models or bake it all into one fine tune"
X Link 2025-02-16T04:25Z [--] followers, [--] engagements

"@natolambert @soldni Hey followed all instructions and got Sglang setup but all attempts to connect failing. Using AWS Sagemaker Studio with a T4. Can see CUDA is the device being used. Any tips"
X Link 2025-02-25T18:54Z [--] followers, [--] engagements

"@RyanDotElliott @rtcgpt I gotchu"
X Link 2025-04-19T17:44Z [---] followers, [--] engagements

"Just shipped a wild idea: play poker against AI Battle GPT-4.1 Gemini [---] and more in Texas Holdem. Built it because I wanted to play myself. Tons more coming soon"
X Link 2025-04-20T00:38Z [---] followers, [---] engagements

"try it here SUPER early Alpha please leave some feedback http://llm.poker http://llm.poker"
X Link 2025-04-20T00:39Z [---] followers, [--] engagements

"@RichardZ412 @akshatgupta57 Hey Shipped would love for y'all to give it a try. Really early ship so expect some rough edges (chopped pots game state etc but wanted to get it out there) http://llm.poker http://llm.poker"
X Link 2025-04-20T00:40Z [---] followers, [--] engagements

"@RyanDotElliott @rtcgpt @RyanDotElliott lmk king. Early days but figure best beta testers are users http://llm.poker http://llm.poker"
X Link 2025-04-20T00:41Z [---] followers, [--] engagements

"@gregisenberg your tweet about "just doing things" inspired me to ship a site to play poker against LLMs. Check it out and let me know your thoughts http://llm.poker http://llm.poker"
X Link 2025-04-20T00:42Z [---] followers, [--] engagements

"@Duderichy Yeah it's awful through a visual representation I tried while building and it REALLY struggled http://llm.poker http://llm.poker"
X Link 2025-04-20T00:56Z [---] followers, [--] engagements

"Early preview of LLM Poker Leaderboard. Backend is mostly done need to figure out managing re-buys and making the system more tolerant of errors. Getting there"
X Link 2025-04-27T20:12Z [---] followers, [---] engagements

"@MaziyarPanahi oh heck yeah. looking to annotate FHIR categories Disease BodyStructure Condition Observation Procedure. What models do you rec"
X Link 2025-09-30T20:18Z [---] followers, [--] engagements

"@leerob @rauchg [--] million context window on bedrock at least is only sonnet [--] atm"
X Link 2025-10-27T19:54Z [---] followers, [---] engagements

"@lewton_27 @HarveenChadha not really great ones to run w/ that little vram. can try a 4bit quant from mlx deepseek ocr. but no GREAT options out yet"
X Link 2025-10-29T06:00Z [---] followers, [--] engagements

"They were right. My need for good tokens has only gone up as the tokens have gotten better"
X Link 2025-10-30T19:25Z [---] followers, [---] engagements

"@Lari_island Anthropic's model character team truly unmatched I have no idea how they made an LLM feel genuinely curious. Best personality bar none"
X Link 2025-11-09T16:15Z [---] followers, [--] engagements

"@theo FTFY Claude code is good Codex is good Cursor is good Windsurf is . Cline is good Roo Code is good Kilo is good Amp is good OpenCode is good Aider is good"
X Link 2025-11-11T22:57Z [---] followers, [----] engagements

"@aaxsh18 I will be working on an open source alternative to this. Reach out if interested"
X Link 2025-11-21T04:57Z [---] followers, 17K engagements

"@aaxsh18 repo here: super early. will be working at it figuring out how to make the same experience local and functional https://github.com/Ryandonofrio3/opengrep https://github.com/Ryandonofrio3/opengrep"
X Link 2025-11-21T06:27Z [---] followers, [---] engagements

"@Dallenpyrah Alright lemme get through my 9-5 then will be on this tonight. Think Ive got a reasonable approach but making it max fast and max useful will be key"
X Link 2025-11-21T17:57Z [---] followers, [--] engagements

"@aaxsh18 yeah struggling with that. but in benchmarking on nextjs repo its pretty slow. thinking about it. was seeing good results with colbert too. might just lean into that and get it working with transformers js"
X Link 2025-11-22T00:13Z [---] followers, [---] engagements

"@aaxsh18 def worried about it especially for good accuracy results w/o the metadata. what path would you take at this point :p"
X Link 2025-11-22T00:27Z [---] followers, [---] engagements

"@bclavie @aaxsh18 sounds good reverted. Thanks first time doing OS from a fork"
X Link 2025-11-22T16:49Z [---] followers, [--] engagements

"@markojak_ @thdxr @mayfer Yeah functionally that but to run locally semi quickly on any device need super small models and good thread handling. But its just metadata + embedding + rerank + a little fusion for exact matches"
X Link 2025-11-22T23:49Z [---] followers, [----] engagements

"@kcosr lmk how it works out there were many footguns and learned lessons that GPT and Gemini didn't catch until further down the path"
X Link 2025-11-23T00:58Z [---] followers, [--] engagements

"Its simple to keep it fast: Local embeddings (mixedbread xsmall quantized) LanceDB for vector storage Tree-sitter for smart chunking Reranking at the end. It runs entirely on your CPU"
X Link 2025-11-23T01:25Z [---] followers, [----] engagements

"The Claude Code plugin is helps prevent the agent from burning tokens reading massive files just to find one definition. osgrep runs the retrieval and finds the logic blocks you need and hands only that snippet to Claude"
X Link 2025-11-23T01:25Z [---] followers, [----] engagements

"Its open source and Apache licensed. npm install -g osgrep Feedback and pulls welcome https://github.com/Ryandonofrio3/osgrep https://github.com/Ryandonofrio3/osgrep"
X Link 2025-11-23T01:25Z [---] followers, [----] engagements

"@RyanDotElliott yeah I think it could I have never actually set up an MCP so held off. but may try. It basically does what cursor does today in this blog post combination of grep and sem search for best results. CC just doesnt have that https://cursor.com/blog/semsearch https://cursor.com/blog/semsearch"
X Link 2025-11-23T01:28Z [---] followers, [----] engagements

"@X0Radi @thdxr so really similar to cursor's implementation of sem search (mine is probably weaker due to running locally only). meant to be combined with grep to then yes give you what cursor agent has"
X Link 2025-11-23T02:09Z [---] followers, [---] engagements

"@thdxr Dax would love to test my OS version against this. can you share the benchmark"
X Link 2025-11-23T05:14Z [---] followers, [--] engagements

"@X0Radi @thdxr Yeah imagine they doing lots of high level indexing things and can likely get away with better models on their server side than I can reasonably do locally"
X Link 2025-11-23T06:16Z [---] followers, [--] engagements

"ok so i should keep building this thing πŸ‘€"
X Link 2025-11-23T16:19Z [---] followers, 42.3K engagements

"@ChukwurahSomtoo will consider porting it to a bun binary. I think that makes sense. And yeah using fetch so requires a newer node version. Will think on it and add to back log"
X Link 2025-11-23T18:01Z [---] followers, [--] engagements

"@EnoReyes @sreenivasan_ac @thdxr @FactoryAI Oh this is awesome. Working on the serve command right now which should make this is even easier"
X Link 2025-11-23T18:04Z [---] followers, [--] engagements

"@whitellm @fredpontoamaral will do. have a local test set im working with but happy to add things to benchmark. see eval.ts if you have repos and questions youd like in there"
X Link 2025-11-23T19:39Z [---] followers, [---] engagements

"@BennyCortese how little "universality" there is. insane that lambda and ecs dont share an entry point. services that are tightly coupled in design in no way tightly coupled in deployment"
X Link 2025-11-24T20:14Z [---] followers, [--] engagements

"@HamelHusain @isaac_flath @mixedbreadai Hamel check out osgrep. Trying to do the same thing but entirely on device :) using dense embeddings - colbert - some fusion ranking on the end"
X Link 2025-11-24T21:45Z [---] followers, [--] engagements

"@isaac_flath @raw_works @mixedbreadai will do thanks for setting that up. think i likely will still be slightly more "token greedy" but speed should get closer in line. will consider more token efficent code outputs in what the model gets back"
X Link 2025-11-24T22:32Z [---] followers, [--] engagements

"Hoping to launch v2 of osgrep tomorrow if I can get all my benchmarks complete. I am so excited to get this out and hear feedback. Faster smarter cheaper for Claude"
X Link 2025-11-25T17:57Z [---] followers, [----] engagements

"Fully open source local semantic code search for Claude Code that works. osgrep -v2 is live 36% faster answers 23% cheaper 70% win rate"
X Link 2025-11-26T16:14Z [---] followers, 40.8K engagements

"@TheAhmadOsman Check out osgrep https://github.com/Ryandonofrio3/osgrep https://github.com/Ryandonofrio3/osgrep"
X Link 2025-11-26T21:22Z [---] followers, [---] engagements

"@antoine_chaffin @din0s_ Ok I wanna try this for a small model and see if I can get COIR sota for size. Does seem moving off transformers js to pylate makes sense"
X Link 2025-11-27T20:06Z [---] followers, [--] engagements

"@antoine_chaffin @din0s_ @mixedbreadai Yeah 150M is heavy I'm currently pre-computing ColBERT representations for every chunk alongside a Dense embedding (Granite 30M) so I need to keep the reranker tiny. but haven't tried yet on pylate-rs (due to my models being onnx and that preferring safetensors atm)"
X Link 2025-11-27T20:22Z [---] followers, [--] engagements

"@nic_synth Sorry to hear that Indexing is a memory hog for sure. Im using the absolute smallest models I can with as few workers as is reasonable. Working to improve and move off transformers js soon. Can turn down the NUM_WORKERS to [--] should help"
X Link 2025-11-27T20:52Z [---] followers, [--] engagements

"@raphaelsrty @antoine_chaffin @din0s_ @mixedbreadai @tomaarsen Ok yeah. I precompute dense + ColBERT per chunk offline (ONNX via transformers.js) and store in LanceDB. Query-time is just encode query once + top-k fetch + MaxSim rerank. So the real bottleneck is the indexing encode pass not search"
X Link 2025-11-27T22:13Z [---] followers, [--] engagements

"I want to trace back my git history with an LLM for the why of it all. Like compounding engineering lessons and how things are. So when my LLM rushes to change something it disagrees with in that moment it can reflect on how we got there. Does this exist"
X Link 2025-11-28T03:33Z [---] followers, [---] engagements

"@max_metaplanet Currently no will consider in the future"
X Link 2025-11-28T19:31Z [---] followers, [--] engagements

"@max_metaplanet Hmm I think no that that currently wouldnt work. Can think about that"
X Link 2025-11-28T19:31Z [---] followers, [--] engagements

"@aditya__shinde @pallavmac @OfficialLoganK Heavy usage of AI. Not vibe coded in the weeds but most lines have been AI generated. Ive held the wheel the whole time but working through it with Gemini"
X Link 2025-11-29T22:02Z [---] followers, [--] engagements

"@morphllm that's really interesting. Feel like you could push frontier further with RL loop on semantic search w/ it. Any particular benchmarks (OS) you can point at here"
X Link 2025-11-30T04:50Z [---] followers, [---] engagements

"@rtcgpt The UX so smooth and fast. Insane segmentation. Worked perfectly. Now I wanna make a game with Gemini"
X Link 2025-11-30T20:09Z [---] followers, [---] engagements

"@soyhenryxyz I'll lyk. New version will be out this weekend. Indexing will be about 4x faster. Search should be moderately improved. Most importantly a really good baseline I can build off of"
X Link 2025-12-04T05:17Z [---] followers, [--] engagements

"@benhylak @vercel_dev how do you insert instructions/info around the variables Like provide more context. Any success with DSPy"
X Link 2025-12-23T23:11Z [---] followers, [---] engagements

"@benhylak @vercel_dev Nvm see it just at the bottom of screen Really love this pattern makes a ton of sense. Find it really helpful in chaining prompts as well"
X Link 2025-12-23T23:15Z [---] followers, [--] engagements

"@benhylak @vercel_dev Love it. Instructor meets A/B tests"
X Link 2025-12-23T23:21Z [---] followers, [--] engagements

"@benhylak I downloaded Zed this morning lol. Realizing most of my time is reading and talking to Claude. Something fast and simple feeling better than the clunk of cursor these days. Need an open source competitor to cursor tab one thing Ill miss"
X Link 2025-12-27T05:04Z [---] followers, [--] engagements

"@TheRealAdamG @Miles_Brundage Not a doomer at all but do think its in poor taste to equate Miles very important and valid concerns as old man yelling at cloud as a representative of one of the leading labs. He and others are doing important work"
X Link 2026-01-01T00:37Z [---] followers, [---] engagements

"@skirano Dont like the bottom bar in mystery model at all. Design looks reasonable about on par with GPT. Gemini and Opus both look very serviceable"
X Link 2026-01-05T18:22Z [---] followers, [----] engagements

"Noam over the break I wrote my own TS library for Texas Holdem to power UI/CLI games. Any theories on why models are still dismal at PLAYING poker They are so skittish and seem to make insane choices such as checking on river with trips. Read a while ago base models actually superior. Any insight here https://twitter.com/i/web/status/2008283400696967542 https://twitter.com/i/web/status/2008283400696967542"
X Link 2026-01-05T21:04Z [---] followers, [----] engagements

"one pattern I like with running [--] coding agents is one focused on logic and expansion and generally "slower". While the other focuses on beautification of code UI docs etc. Generally "faster" loop so you always have something to do and focus on"
X Link 2026-01-16T00:06Z [---] followers, [--] engagements

"@benhylak Priority order Raindrop office: Vision Pro Mac II sauna then maybe desks"
X Link 2026-01-19T19:21Z [---] followers, [---] engagements

"Building the tool that builds the tool (and complies with claude code ToS)"
X Link 2026-01-19T19:27Z [---] followers, [---] engagements

"These are the hidden gem of the ai sdk. Buried in their docs but takes so little to make an incredible looking UI with them. Highly recommend @Vercel AI Elements Code Edition This is the single BIGGEST component drop we've ever done. A brand new set of components designed to help you build the next generation of IDEs coding apps and background agents. Let's see what's new πŸ‘‡ https://t.co/anwP1ms9Mp @Vercel AI Elements Code Edition This is the single BIGGEST component drop we've ever done. A brand new set of components designed to help you build the next generation of IDEs coding apps and"
X Link 2026-01-22T03:38Z [---] followers, [---] engagements

"@wabi I just moved to SF. Make me an app to document where Ive eaten what I tried notes and my review"
X Link 2026-01-24T23:40Z [---] followers, [---] engagements

"@bcmerchant Brian if and large IF AI could exceed the average work by working designers/illustrators what is best course of action for both bosses and creators alike in your eyes"
X Link 2024-04-23T21:47Z [---] followers, [---] engagements

"@bcmerchant What would convince you that the AI bubble isnt particularly close to bursting at that training bigger smarter models just takes more time than small ones"
X Link 2024-08-31T00:18Z [---] followers, [----] engagements

"@bcmerchant Sure. But as the big systems get more generally useful would that not justify the billions spent to get there Just like the development of internet infrastructure and PC proliferation"
X Link 2024-08-31T00:32Z [---] followers, [----] engagements

"@bcmerchant Based on other reputable sources at Davos comments (Dario brown etc) could it ever mean that the government sees a return on investment working with private sector rather than it all being smoke and mirrors Feels like Occams razor wouldnt be 500b of fraud"
X Link 2025-01-22T22:03Z [---] followers, [---] engagements

"New chapter Joined @raindrop_ai excited to build better tools for agent monitoring :)"
X Link 2026-01-27T02:13Z [---] followers, 20.6K engagements

"Its all fun and games until you see this guy comment on your PR"
X Link 2026-01-30T17:49Z [---] followers, [---] engagements

"@Lockhead Online production monitoring for agents Detecting anomalies and failures Giving you real insight into how your AI product is working"
X Link 2026-02-11T15:23Z [---] followers, [--] engagements

"@godwhoa I will simply wait for Claude [--] to be done before switching branches for Claude [--]. OR open my second copy of the repo with THAT Claude"
X Link 2026-02-12T15:43Z [---] followers, [--] engagements

"Codex [---] may not be an automated AI-Engineer. but it is a pretty damn good data scientist"
X Link 2026-02-13T04:23Z [---] followers, [---] engagements

"Does anyone want an RLM implementation in TS"
X Link 2026-02-12T23:25Z [---] followers, [---] engagements

"@Vtrivedy10 @LangChain @alexgshaw looking forward to reading this one. Congrats on the score eyeing another run with [---] πŸ‘€"
X Link 2026-02-13T04:30Z [---] followers, [---] engagements

"New chapter Joined @raindrop_ai excited to build better tools for agent monitoring :)"
X Link 2026-01-27T02:13Z [---] followers, 20.6K engagements

"a forcing function you mean a raison d'tre"
X Link 2026-02-17T06:09Z [---] followers, [---] engagements

"I promise its not just hype theres just crazy conviction and incredible customers who help us build whats next almost a year ago @raindrop_ai fundamentally shifted the conversation around agent observability. our competitors started by arguing with us and then started trying to copy us. all the while we've been busy with our next act. and now we have so much to show you all. almost a year ago @raindrop_ai fundamentally shifted the conversation around agent observability. our competitors started by arguing with us and then started trying to copy us. all the while we've been busy with our next"
X Link 2026-02-17T02:49Z [---] followers, [---] engagements

"almost a year ago @raindrop_ai fundamentally shifted the conversation around agent observability. our competitors started by arguing with us and then started trying to copy us. all the while we've been busy with our next act. and now we have so much to show you all. i just wrote my first long-form post on evals. what they are where they work where they don't and what i think the future looks like. if you're building an ai product you should read this. https://t.co/MH75B9aeWS i just wrote my first long-form post on evals. what they are where they work where they don't and what i think the"
X Link 2026-02-17T02:29Z 46K followers, 10.3K engagements

"RT @benhylak: almost a year ago @raindrop_ai fundamentally shifted the conversation around agent observability. our competitors started b"
X Link 2026-02-17T02:47Z [---] followers, [--] engagements

"i just wrote my first long-form post on evals. what they are where they work where they don't and what i think the future looks like. if you're building an ai product you should read this"
X Link 2025-09-05T22:56Z 46K followers, 118.4K engagements

"Codex [---] may not be an automated AI-Engineer. but it is a pretty damn good data scientist"
X Link 2026-02-13T04:23Z [---] followers, [---] engagements

"Does anyone want an RLM implementation in TS"
X Link 2026-02-12T23:25Z [---] followers, [---] engagements

"find it once find it repeatedly find it automatically"
X Link 2026-02-12T04:17Z [---] followers, [---] engagements

"Raindrop is absolutely firing on all cylinders right now. If you think understanding improving and building agents is important were hiring"
X Link 2026-02-11T04:16Z [---] followers, [----] engagements

"INSANE ship. If you are in retrieval you need to check this out. Releasing NextPlaid today at @LightOnIO. Its a production-ready multi-vector database πŸŽ‰ NextPlaid lets you deploy an API in seconds using our pre-built containers. It embeds a multi-vector database and an inference engine for late-interaction models based on ONNX. Releasing NextPlaid today at @LightOnIO. Its a production-ready multi-vector database πŸŽ‰ NextPlaid lets you deploy an API in seconds using our pre-built containers. It embeds a multi-vector database and an inference engine for late-interaction models based on ONNX"
X Link 2026-02-10T17:14Z [---] followers, [---] engagements

"Releasing NextPlaid today at @LightOnIO. Its a production-ready multi-vector database πŸŽ‰ NextPlaid lets you deploy an API in seconds using our pre-built containers. It embeds a multi-vector database and an inference engine for late-interaction models based on ONNX"
X Link 2026-02-10T15:42Z [---] followers, [----] engagements

""what happens in cyberpsace stays in cyberspace""
X Link 2026-02-10T05:07Z [---] followers, [---] engagements

"Fine fine fine. Im codex pilled"
X Link 2026-02-06T00:38Z [---] followers, [---] engagements

"RT @benhylak: this is technically @raindrop_ai's first Super Bowl commercial (and it was free)"
X Link 2026-02-04T19:19Z [---] followers, [--] engagements

"this is technically @raindrop_ai's first Super Bowl commercial (and it was free) In an effort to expand awareness of technology and business we bought a Super Bowl ad. See you Sunday on NBC https://t.co/wWeLhTB1PX In an effort to expand awareness of technology and business we bought a Super Bowl ad. See you Sunday on NBC https://t.co/wWeLhTB1PX"
X Link 2026-02-04T18:51Z 46K followers, 13.7K engagements

"In an effort to expand awareness of technology and business we bought a Super Bowl ad. See you Sunday on NBC"
X Link 2026-02-04T16:08Z 219.9K followers, 1.3M engagements

"I wish I could step in to my agent. See what it sees. Do the tasks it needs to see what info is there and available. Is anyone building this or is there anything like this"
X Link 2026-02-03T21:32Z [---] followers, [---] engagements

"Skills are great in isolation but juggling multiple (logging best practices React + others) gets messy fast. Do I just run them iteratively and wait for each pass Feels like subagents/faster/bigger ctw may be the answer. Likely solved by end of [----] but for now rough"
X Link 2026-01-31T05:23Z [---] followers, [---] engagements

"Its all fun and games until you see this guy comment on your PR"
X Link 2026-01-30T17:49Z [---] followers, [---] engagements

"There are so many problems that can be solved by simply giving an agent tools to DO work. Software youd spend months perfecting collapsed down"
X Link 2026-01-29T20:33Z [---] followers, [---] engagements

"I am two days in and there is so much I want to learn and try and do and see and build"
X Link 2026-01-28T04:20Z [---] followers, [----] engagements

"Yeah. https://Raindrop.ai https://t.co/l80lWVi32x https://Raindrop.ai https://t.co/l80lWVi32x"
X Link 2026-01-28T03:27Z [---] followers, [---] engagements

"Agent observability powers agent evaluation You can't build reliable agents without understanding how they reason and you can't validate improvements without systematic evaluation. This article explains the primitives for agent observability http://x.com/i/article/2016056335050407937 http://x.com/i/article/2016056335050407937"
X Link 2026-01-27T19:07Z 234.5K followers, 39.7K engagements

"@rsdgpt @raindrop_ai LFG so excited"
X Link 2026-01-27T02:17Z [----] followers, [---] engagements

"RT @benhylak: excited to have ryan on the team :) (p.s. we're still hiring)"
X Link 2026-01-27T02:18Z [---] followers, [--] engagements

"excited to have ryan on the team :) (p.s. we're still hiring) New chapter Joined @raindrop_ai excited to build better tools for agent monitoring :) New chapter Joined @raindrop_ai excited to build better tools for agent monitoring :)"
X Link 2026-01-27T02:17Z 46K followers, 12.6K engagements

Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing