Dylan Huang

@ethanelasky DMs open!

@dphuang2 can you open dms, I have some questions on future tinker capabilities and don't wanna invest in a lot of infra if you guys are gonna put more infra out soon

this is actually crazy

@soumithchintala I can't tell if this is AI or not, I even googled it 😅

someone's getting started early!

This week’s projects expand the horizons of training in Tinker, from massive libraries of RL environments to new methods of training agents and models that budget their own thinking.

@karinanguyen congrats on releasing PostTrainBench! Have you guys thought about using Tinker in the harness? I imagine it would remove some of the setup/infra debugging work the agents need to do.

Excited to release PostTrainBench v1.0! This benchmark evaluates the ability of frontier AI agents to post-train language models in a simplified setting. We believe this is a first step toward tracking progress in recursive self-improvement 🧵:

We are partnering with @nvidia to power our frontier model training and platforms delivering customizable AI. thinkingmachines.ai/news/nvidia-pa…

@a_levitator what do the internal numbers say

Ramp inspect might be the greatest internal tool any tech company has ever made

@HanchungLee I think startups that don’t figure this out might become undifferentiated, and ultimately die

the model is the product.

@alvinsng @FactoryAI engineers should be spending more time allowing agents to iteratively verify their work 🙂‍↕️

At @FactoryAI, every PR triggers 40+ CI checks, all finishing in under 6 minutes. Our automated guardrails are so fast and comprehensive that you can "merge recklessly". This is agent-native development

Contextual AI used Tinker to post-train the planning behavior for a search agent. They land on a two-stage training recipe: On-Policy Distillation and GRPO with a CLP reward. Read more 👇

@xn1cklas @QuiverAI @joanrod_ai Can’t wait to try it out in more projects 🙌

@dphuang2 @QuiverAI @joanrod_ai wow this is really cool, especially for such a "simple" prompt!

@QuiverAI is amazing... These logos took a couple minutes to create. I've played around with frontier models to generate SVGs, nothing compares. Great job @joanrod_ai and team! Looking forward to Arrow-2.0.

Introducing @QuiverAI, a new AI lab and product company focused on frontier vector design. We’ve raised an $8.3M seed round led by @a16z, with support from amazing angels and investors. Our first model, Arrow-1.0, generates SVGs from images and text. It’s available now in public beta at app.quiver.ai

@noahzweben new attack vector acquired

Announcing a new Claude Code feature: Remote Control. It's rolling out now to Max users in research preview. Try it with /remote-control Start local sessions from the terminal, then continue them from your phone. Take a walk, see the sun, walk your dog without losing your flow.

@stephenhaney why not just have agents directly write code and hotload the results in the browser? Is the Paper DSL less confusing for agents? Is it a better workflow for designers? Just curious about your insights.

Hello! Today we're releasing Paper Desktop Paper is now a canvas for Cursor, Claude Code, Codex. Any agent can read and write html to Paper. • push or pull from your codebase • pull real data from anywhere • less work, more design What will you ship? Sound on 🎶

@jaltma AI infra or APIs that agents use

The current consensus view is saas is dead...presuming that's right, the next interesting next question is What companies are "safe from ai"? - handling money, regulation - agents on top of company data - most hardware? - maybe systems of record? - security? - marketplaces?

@VihaarNandigala Congrats!!

We just raised a $5.3M seed round for Orange Slice, co-led by 1984 Capital and Moxxie Ventures, with participation from angels like Paul Graham. We’re building AI agents, inside a spreadsheet, that help sales teams find companies that already want to buy. The reality is most sales teams don’t struggle with effort - they struggle with timing. Reps spend huge amounts of time working static lists and broad targeting, chasing leads that were never going to convert. That creates noise, low reply rates, and wasted cycles. Top companies like Ramp solve this with dedicated growth engineers building internal data workflows. We’re making that same capability accessible to everyone else. At its core, the challenge is simple: finding customers who already have the problem you solve. Orange Slice turns the spreadsheet into a system for discovering buying signals - agents research company sites, news, social signals, and niche sources like court records or building permits, then structure that information directly into columns teams can act on. Not “might be a fit.” But “likely in-market.” So instead of guessing who to target, teams build and refine living lists of high-intent accounts inside a sheet. Still early. Still learning. But we’re excited to keep building. Kishan and I met sophomore year on a Bollywood dance team at Michigan — and I couldn’t ask for a better co-founder. Grateful to our team, customers, and investors for believing in this vision.

@shensi @merge_api @t_patterson @prit4k LMAO

It takes years to build a reputation and seconds to destroy it. And last night, we destroyed @merge_api's. Christian McCaffrey and Olivia Culpo sat next to our exec team during our board dinner. Our CRO @t_patterson is a huge football fan, and asked @prit4k to sneak a picture for him. Pritak did.. but he left his flash on. Even worse, this is the picture he took.

🌕 Kimi K2.5 = open SOTA reasoning + vision + 256K context + agentic coding 🏎 200+ t/s on @FireworksAI_HQ (soon even faster) ✅ Nails @simonw's "pelican on a bike" test in both directions Try it now on Fireworks and hats off to @Kimi_Moonshot