Dror Katzav

Thrilled to announce that Together Fund is investing in Sentra, alongside a16z speedrun! You track results in Jira. Decisions in Notion. Conversations in Slack. But the reasoning, the debates, trade-offs, and context behind why you chose A over B, disappears into what we call "Dark Matter." A decision made in March looks insane by July because no one remembers the constraints that made it smart. I lived this firsthand at Twitter scaling from 800 to 8,000 employees, and at Google while launching AI Overviews to billions at planet scale. The problem isn't process. Process is compensation for something deeper: organizational amnesia. An organization’s "Systems of Record" doesn’t solve this, they encode it. They store what happened, never why. That's why we are investing in Sentra. Sentra is the always-on collective memory that eliminates organizational amnesia by maintaining accurate context for all members and agents, functioning as an operational nervous system. It connects to every channel where work happens, meetings, Slack, email, code commits, docs, calendars, and treats them not as artifacts to search, but as living signals to synthesize. The fleeting and the permanent, unified into a memory that understands. The founding team is built for this: - Jae Gwan Park (CEO): Product-first founder, memory systems research at UofT and MIT - Ashwin Gopinath (CSO): Former MIT professor, created "Reflexion" (NeurIPS 2023), agents that learn from mistakes, 2x founder - Andrey Starenky (CTO): Early Vapi engineer, ex-IBM, built to process enterprise-scale data firehose Together is an operator-led fund. We invest in problems we've lived. This is one of them. Many congrats Jae, Ashwin and Andrey, we are so excited to partner with you! Read the full thesis: https://lnkd.in/gixj9cE4 Book a demo: https://www.sentra.app/ #OrganizationalMemory #AI #Sentra #TogetherFund #a16z #ContextGraphs

71

3 Comments

AI is changing what “scale” actually means. In CTech by Calcalist’s 2026 VC Survey, Haim Bachar and Lee Ben-Gal share how we’re seeing the scaling metric shift - from headcount growth to output per person. This isn’t about doing more with less. It’s about building companies differently from day one - and rethinking how success is measured. See the full article below 👇

13

Last month, in Y Combinator's call for startups for their Spring 2026 cohort, they called for "AI Native Agencies." YCs frames this opportunity simply: "agencies of the future will look more like software companies, with software margins. And they'll scale far bigger than any agencies that exist in these fragmented markets today.” We believe YC is right. We differ, though, with YC in our belief in where the value will accrue. More on this here 👇👇 https://lnkd.in/g6iNtTUh

21

5 Comments

A lot of VCs use founder age as a proxy for how well they’ll leverage LLMs and the latest AI. In my experience, the stronger signal is how technical they are. Technical founders understand how LLMs work and can exploit them fully; non-technical founders see magic, not mechanisms.

49

11 Comments

With RL techniques like GRPO, it’s become easy to fine-tune small models to outperform foundation models on vertical tasks. We’re open sourcing Osmosis-Apply-1.7B: a small model that merges code (similar to Cursor’s instant apply) better than foundation models. Links in comments! Using foundation models for high specificity, low complexity tasks like code merging can be overkill - instead, it’s better to use a specialized model instead for higher performance and latency. We fine-tuned the model on 100K real-world code commits where the model needs to merge a code edit snippet into an original code snippet. On the test dataset (subset of CommitPackFT), foundation model performance ranged between 0.77-0.93 reward score. In comparison, Osmosis-Apply-1.7B achieved a 0.98 reward score - while also being 3X-5X more affordable than the next cheaper model AND ~10X faster! Why this matters: the best AI agents use multiple models to ensure task specialization and reliability at scale - on top of cost and latency benefits as well. Links to the model and full write-up (with reward function, hyperparameters, etc.) are in the comments - reach out if you’re interested in using reinforcement learning!

240

18 Comments

The big labs are expansive in their product ambition, especially since foundation models have largely improved in lockstep - in order to compete with them you have to do things they won’t which are: - building a very rich software ecosystem around a primitive - orchestration across multiple models - going insanely deep on product and growth for a narrow vertical domain

58

3 Comments

I had a great time chatting with Patrick O'Shaughnessy on Invest Like The Best. I've known Patrick since college, and this is the first time we've talked markets and investing at this much depth. The fundamentals of company building haven’t changed: people, products, and markets matter. But obviously, private markets have evolved substantially over my career: there are now ~6x more private unicorns than public companies with a $1b+ market cap. And at the end of 2010, just 2 public technology companies were among the top 10 in market cap; today it’s 8 of 10. AI (alongside software eating everything more generally) is clearly driving a lot of this. But it’s instructive to look at everything from the steam engine, to the early days of Facebook and Google user monetization, to real-time success stories like Databricks, Anduril, OpenAI and Waymo, to get a clear picture of where the opportunities lie. It was a pleasure to go deep on all this and more!

131

9 Comments

SF VC Takeaway #3: Early velocity is getting companies funded, but durability will decide who lasts. If the last couple years of venture rewarded velocity, many investors are already preparing for a shift back to durability as the early AI experimentation phase starts to come to an end. Right now, the market is still funding: fast revenue ramps, strong early momentum, and compelling AI-native narratives, but there’s a shared understanding that this is a unique moment in time. Series A rounds are happening so quickly after seed that many companies are raising rounds today without having to answer questions like: How durable is this revenue? What does retention look like after the initial excitement fades? Where do gross margins settle once pricing, usage, support, and compute costs normalize? Investors know they’re making investment decisions before these questions are fully answered, but they are still mindful that in the coming years the market and economics of these businesses will settle and mature. They can't miss the wave of current innovation, but that means they are investing with a higher degree of uncertainty. They are eyes wide open about that. Several people framed it the same way: “we’re underwriting velocity now, but watching durability very closely.” Over the next 6–18 months, many investors believe the market will transition from experimentation to expectation and from “how fast did this grow?” to “does this actually hold up?” When that happens, the questions that start to matter more are less about top-line speed and more about: 🏋🏽 Quality of revenue 🔁 Retention through renewal cycles 💸 Long-term gross margin structure This is especially relevant for AI-native companies where: engineering and GTM costs are more fluid, services (ahem forward deployed engineers) are often used as a wedge, and margins can either look great or terrible early on until scale changes their structures. The current market assumes the winners stay winners (often upheld by capital moats), but many investors are already watching closely to see which companies still look strong once the experimentation phase ends. None of this means velocity stops mattering, it just means companies will have a higher bar to raise mega-funding and will have to be both durable and high growth. My take (even though I promised no opinions) is that velocity matters right now, but in the not too far off future (6-12 months), the market will reward the right balance of velocity with durability and some of the more methodical companies will prevail in the long run. I think VCs will rather fund the high quality $2-3M ARR business than the uneasy $4-5M ARR business, but time will tell. That’s my final SF VC takeaway for the week. Next week, I’ll shift to what I learned from operators actually building AI-native companies and how scaling a company today looks much different than even just 1-2 years ago. #venturecapital #SFVCTakeaway

40

3 Comments

Replit just raised $400 million at a $9 billion valuation, and shipped its fourth autonomous coding agent in two years. What started as a tool for professional developers in 2016 is now pioneering how anyone — regardless of technical background — can build software. The story behind how that transition actually happened is less told. That’s a design story — and the domain of Haya Odeh, its co-founder and VP of Design. Valentina Suarez Gutierrez, Yina L. and I recently sat down with Haya to talk about the design decisions behind making software creation feel intuitive for people who’ve never written a line of code. (Some changes are as small as a single word.) She also shares her path into design, an inside look at how Replit's design team operates, and why taste — the kind that notices the difference between 1% gray and white — is something that every designer still needs, especially in the age of AI. https://lnkd.in/gu6QTCSf

52

4 Comments

This a seductive idea; that organization's context fragments, decision logic gets lost, and an organization is prone to tackle collective amnesia. BUT this comes with a lot of caveats!! - MONOCULTURE: When one system becomes the authoritative interpreter of "why things are the way they are", it will collapse pluralism. An organization actually benefits from various "contested memories". An engineer and sales guy remember different things from a discussion and that helps. Employees can stop arguing with the system not because its correct but its "coherent". - OBSERVER's EFFECT: When people know that the reasoning within systems is being captured, they will change what gets recorded. The negotiations will become "safer" and will contaminate raw signals. - CENTRALIZATION: Interpretation gets centralized upwards. It may appear leadership or executives are gaining synthesis and becoming closer to the ground. But there is no comparison to living through the frictions, not even eclectic summaries. LEADERSHIP MUSCLE NEEDS TO PRACTICE LIVING THROUGH FRICTION OTHERWISE IT WILL TURN SENILE. - BIAS AMPLIFICATION: Model's inbuilt biases can combine with company's cultural quirks resulting in misinterpretations and distortions, worse they get reinforced over time. System will learn how the company thinks and then will optimize for continuity. AMNESIA IS SOMETIME A FEATURE. - SEMANTIC TRUTHS: Unlike objective truths this system will play on subjectivity and will build "causal" relationships that are based on semantics. This over time can form mythical decision norms with no external validity. The real question isn't "CAN IT BE BUILT?". Of course it can be. But "HOW IT SHOULD BE POSITIONSED". #llms #leadership #ai

1

Super exciting to see reports this morning that Waymo is in advanced discussions to raise $10–15B at a valuation near or above $100B, with Alphabet expected to anchor the round. That would be a major step up from Waymo’s prior ~$45B valuation in 2024 and a clear step change in investor confidence. Notably, Waymo has not raised this year, despite the broader AI funding frenzy. The size and structure of the round also make this feel like a step toward an eventual spinout or IPO. This matters because capital supercharges growth. Waymo is past the “does this work?” phase and firmly in “how fast can we scale?” mode. The company has already logged ~14M rides this year and expects to reach ~1M rides per week by late 2026 as it expands into markets like Miami, Dallas, and Philadelphia, including expanding onto freeways. At the same time, competition and consumer expectations are accelerating. In markets where Waymo operates, fully driverless rides are becoming normal rather than novel. Once people can reliably hail a car with no one in the driver’s seat, tolerance for “pilot programs” and distant timelines drops quickly. The race is no longer about whether robotaxis work. It is about who can scale them first, safely, and everywhere — a win for consumers who get cheaper, safer, and more reliable rides.

25

While there has never been a more exciting time to be a founder building dev tooling or next-gen infra, it has also never been less investable at seed/pre-seed. I'm either really missing something or a lot of my peers are lost. As someone who has not just written, but also SHIPPED, about 75k lines of code in the past 6 months I can tell you that the evolution of how to build products has changed as much in the past year as it did in the entirety of 2007-2017. The complete rise and fail of frameworks, platforms, methodologies, etc... paved over and forgotten... that is of course except for the 1 company that gets a 1000x return from a wildly overvalued hyper-scaler or drunken growth stage investor obsessed with compounding at scale. Imagine a world where any seed investor in trends like Openstack, Hadoop, PaaS, etc all took a full loss on their investment. That's what we're looking at right now. I personally know of over a dozen well-funded seed-stage companies building in these spaces, with years of runway, scrambling to get acquired for a return of capital + several million personally while they're still relevant. If you're not seeing this unfold in front of you, you either aren't paying attention or you're satisfied playing the lottery instead of investing.

75

11 Comments