Guy Hartstein

imo, the most important and interesting problem to work on today is organizing a 'company environment' for your agents to operate in. Concretely - this means creating a filesystem/repo that models your entire company...cc/codex/cursor will live and operate in this repo.

@tom_doerr 🫡

Multi-agent tool for automated company research github.com/guy-hartstein/…

@bwein_ @SynchronicityHQ 🐐

Modern markets are fundamentally extractive We've spent three years rewiring them from scratch One core feature is everyone deploying their own bots without code or infra Here's a first look @SynchronicityHQ We're looking for strategists to test drive. Reply for early access

@shcallaway 🍻

Nebius, the GPU cloud, acquires Tavily, the agentic web search company. We were early users of Tavily at 11x. Their APIs were a big part of what allowed us to make an AI sales rep that felt truly "human" (she would pull in relevant, real-time info from all over the web for her email campaigns). It sounds a lot less remarkable today, but in late 2024 it was revolutionary. Tavily defined this category in the early days and is continuing to push it forward. Congrats to @rweiss57 and the @tavilyai team!

@tavilyai @nebiusai Amazing day for an amazing team 🔥

We’re joining @nebiusai 🚀 What’s changing: Tavily now has the backing of a global AI cloud, meaning faster global latency, higher uptime, and a clear path to SOC2/HIPAA for enterprise agents. What’s not changing: Tavily remains a standalone, cloud-agnostic API. Same keys, pricing, support, and the same team building the web for AI agents just with a much bigger engine. More soon.

one thing i’ve observed is that the companies that add tremendous value today excel at applying AI while simultaneously developing infrastructure that’s loathsome to build. AI web search is a great example. Building great infra takes up 98% of developer time, yet the AI aspect drives 98% of the value. Same with agents for enterprise. AI reasoning and tool use drive the visible value, yet 90% of the work is permissions, sandboxing, safeguards, integration with legacy systems etc.

@deansacoransky grep pilled

file system pilled

NYC AI Builders! 🛠️ 🗽 Want to see how @TavilyAI hit the top of the DeepResearch Benchmark? 📈 Join us with @datadoghq & @daytonaio as one of our engineers breaks down their new SOTA Research Endpoint. See you Thursday! 👇 luma.com/gultpezn

@tavilyai @datadoghq @daytonaio bullish on @yash_alap

Slop is a kind of digital kitsch. It’s naïve imitation (perhaps inevitably due to token prediction); a sanitized view of the world that demands to be taken seriously. Yes, it’s low signal, but the reason we have found a name for it is because it’s a familiar aesthetic. Milan Kundera puts this quite well: “Kitsch causes two tears to flow in quick succession. The first tear says: How nice to see children running on the grass! The second tear says: How nice to be moved, together with all mankind, by children running on the grass! It is the second tear that makes kitsch kitsch.” I’m not sure if you can prompt against this. When you distill the world’s information, you get eccentric, earnest imitation that creates the formulaic, sentimental aesthetics that avoid confronting complexity and meaninglessness.

Has anyone encountered a good definition of “slop”. In a quantitative, measurable sense. My brain has an intuitive “slop index” I can ~reliably estimate, but I’m not sure how to define it. I have some bad ideas that involve the use of LLM miniseries and thinking token budgets.

@willccbb I like that model companies nowadays aren’t just producing all-purpose SOTA models (like generic cars, built for everyone) but instead are rapidly iterating on small, capital efficient models that excel at a small subset of tasks. Why build a car when you can build a jetpack?

starting a model company now seems kinda like starting a car company. in the early days, it required insane amounts of R&D + risk. but now, we basically know how to build cars. put enough smart people in a room and let them hammer at it for a while and you’ll get a car

@SnowballHustle @DPazdan @Polymarket @NTmoney There would have been significant dilution from A onwards. That $50k would be much closer to 1m than 4m unless he followed on with each round to maintain ownership.

@DPazdan @Polymarket @NTmoney that $50k would have been close to $4M today right?

Had the chance to join @Polymarket's bridge round @ $120M val… but passed🫠

@benln @tavilyai 👀

Very few opportunities to join a breakout startup as it's taking off @tavilyai is dominating in the AI search space, hiring for a ton of roles

@benklieger @GroqInc @4rnavJ @marmikch Fantastic ☄️

Excited to be announcing the next generation of Compound, @GroqInc’s first agentic system! This the result of a lot of hard work, including from @4rnavJ and @marmikch who joined us for the summer! Thrilled to be leading this work and scaling Compound 🔥

@MartinShkreli Why is the peer group constrained to shareholders? In theory, the consensus is also the product of capital allocators that *didn’t* invest, creating an implicit evaluation

“consensus” means nothing, really, without defining the peer group. Every company’s shares are 100% owned by its shareholders: so it’s consensus to that finite group and non-consensus to literally every other investor that participates in that asset class. If you define it as T1VC…

@joshvancuyck Incredible work @joshvancuyck

Quite a milestone! Very proud of how far this project has come. Scope the knowledge hub for some interesting tid bits: veritas.computer/knowledge-hub

Introducing the Veritas whitepaper: veritas.computer/veritas.pdf The internet has no native mechanism for discovering what’s true. Veritas is a protocol for humanity to discover what it truly believes - even when it is subjective, uncertain, and changing.

@veritas_layer Fantastic!