Joined November 2025
954 Photos and videos
Pinned Tweet
🚨 @karpathy followed @thdxr @thdxr is building OpenCode (@opencode), an open-source coding agent at @anomalyco
16
27
1,166
295,816
πŸ—žοΈ @ClementDelangue followed @eisokant Eiso Kant co leads @poolsideai to create AI tools that generate and optimize code. Recent posts also show him leaning harder into the broader argument about AI, labor and power.
3
103
πŸ—žοΈ @karpathy followed @TheZvi Zvi Mowshowitz covers AI capabilities and safety topics at "Don't Worry About the Vase" and established Balsa Research to advance evidence based policy interventions.
1
5
136
πŸ—žοΈ @poolio followed @TheZvi
1
24
πŸ—žοΈ @rauchg followed @marvinvonhagen Marvin von Hagen co-founded Poke (@interaction), an AI agent built for the inbox and messaging layer rather than another standalone chat app. Apple approving Poke for Messages gives the product a real distribution edge.
14
634
πŸ—žοΈ @swyx followed @0xluffy @0xluffy founded @capydotai to enable conversational threads that reference and build on each other directly in AI workflows.
6
133
πŸ—žοΈ @bryan_johnson followed @markiewagner @markiewagner is building @PoeticHQ, with a focus on AI that can carry longer workflows without the usual agent flakiness. The launch pitch was pretty specific multi hour execution, high accuracy and lower token burn.
Introducing @PoeticHQ: a new AI system that executes complex multi-hour tasks with 99% accuracy and 10x fewer tokens than agents. We raised $50M at $500M from Kleiner Perkins, Founders Fund, First Harmonic, and Genius Ventures to build AI that does complex work inside Fortune 500 companies without hallucination. While code is too brittle, agents are too unpredictable. The work that runs the global economy - anti-money laundering, fraud investigations, underwriting - needs extreme accuracy. So we built a new kind of software that pairs the flexibility of AI with the predictability of code. When the world stays the same, Poetic runs fixed code: fast, cheap, identical every time. When the world changes, Poetic uses AI to regenerate its approach and find its way back to the objective. In one year, we went from zero to an eight-figure run rate as a team of four. Since then, we’ve scaled the team and executed the highest-stakes processes at AIG, SoFi, and Chime. At SoFi, a large US bank, Poetic reached 99% quality on fraud investigations in five weeks.
4
121
πŸ—žοΈ @jerryjliu0 followed @hxiao @hxiao leads AI at @elastic after founding @JinaAI_. Recent work extracts knowledge graphs from private corpora with Qwen3 models on L4 GPUs to generate verifiable multi hop questions for agentic search eval sets.
3
118
πŸ—žοΈ @karpathy followed @petergostev @petergostev is the @arena researcher behind BullshitBench and a lot of recent work on model reliability. Current activity includes analysis of agent sub threads, model constraints in Fable and real time world model experiments.
5
119
πŸ—žοΈ @dkhos followed @dadiomov @dadiomov leads @ModernTreasury as president after its YC S18 origins. The platform handles high volume financial workflows that now intersect with tokenized asset markets and real estate financing automation.
2
73
πŸ—žοΈ@rauchg followed @adilmania @adilmania has been turning @siliconmania into a new kind of tech media feed, with weekly recaps, sharper founder interviews and late night show energy instead of the usual two hour podcast drag.
5
157
πŸ—žοΈ @drfeifei followed @hugothomel @hugothomel leads @alakazam_ai backed by @fdotinc. The platform turns any sentence into a living game via real-time world models with interactive characters, dynamic rules and custom goals all handled by AI with no code required.
Jun 10
we brought a game back from the dead with a world model Anthem got killed in January when its servers shut down today Freelancers can fly again no game ever has to stay dead again! try it for a limited time on play.alakazam.gg
4
168
πŸ—žοΈ @denny_zhou followed @Apodex_AI @Apodex_AI delivers SOTA deep research with Apodex 1.0. The heavy-duty agent team runs parallel sub-agents for web exploration then applies global verification before committing any answer.πŸ‘‡
Meet π—”π—½π—Όπ—±π—²π˜… 𝟭.𝟬 πŸ”­ β€” a heavy-duty agent team for deep research, which sets the SOTA! The team searches the web, reasons over evidence, and writes reports where every claim is backed by an explicit 𝘦𝘷π˜ͺπ˜₯𝘦𝘯𝘀𝘦 𝘀𝘩𝘒π˜ͺ𝘯, independently audited before delivery. 🌐 apodex.ai
4
224
πŸ—žοΈ @ClementDelangue followed @latkins @latkins leads as CTO of @arcee_ai, the startup behind Trinity Large, a 400B-parameter open foundation model trained from scratch in record time. The company emphasizes reliable, efficient AI designed for real enterprise scale on any infra.
Today we're releasing Trinity-Large-Thinking. Available now on the Arcee API, with open weights on Hugging Face under Apache 2.0. We built it for developers and enterprises that want models they can inspect, post-train, host, distill, and own.
2
161
πŸ—žοΈ @soumithchintala followed @KaichaoYou and @inferact @KaichaoYou is the Inferact cofounder and vLLM maintainer helping turn one of the most widely used OS inference engines into a company. @inferact launched this year with $150M to make model serving cheaper and faster.
6
136
πŸ—žοΈ @steipete followed @timshi_ai @timshi_ai is a @Recursive_SI cofounder, now working on AI that automates knowledge discovery instead of just shipping another app layer. The company came out of stealth recently with $650M behind that bet.
4
217
πŸ—žοΈ @soumithchintala followed @TRReardon @TRReardon is building @flourishailabs after years in neural interfaces and neuroscience. The new angle is not another model lab, but using cortical learning principles to make AI less power hungry and better at continuous learning.
4
97
πŸ—žοΈ @swyx followed @Scobleizer @Scobleizer has been turning his long-running tech curation into Aligned News, an AI-written news product built from 63 X lists and 100,000 plus accounts. Lately he has been especially focused on AI, agents and spatial computing.
Much improved AI news from X: alignednews.com/ai My agents have been busy reading all of you. And I've been busy yelling at them to do better. Well, OK, showing them tons of stuff they missed and building new algorithms to find better stuff. Someone told me "when you have AI reading X you'll have so much extra time." Liar. Now I still read X, but having this agent means I am pouring hours into making my lists better, to make it better, and then I see so many ways it could be better, forcing me to learn how to educate it better. Meanwhile DMs and emails are piling up. Tomorrow will be in SF at the Upscale conference. Bringing my laptop so I can catch up.
3
126
πŸ—žοΈ @sama followed @ericho_goodfire @ericho_goodfire is building @GoodfireAI, the interpretability startup trying to make AI models easier to understand, debug and steer. Goodfire has been pushing that case hard with its Silico platform and a recent $150M Series B.
3
130
πŸ—žοΈ @rauchg followed @alighodsi @alighodsi is steering @databricks deeper into agent infrastructure, with recent work around Lakebase, Agent Bricks and security tooling like Lakewatch as the company chases the enterprise AI stack end to end.
4
124