Tomer Kagan

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

CRV Security: Request for Startups I never know if this actually works for our friends over at YC but figured we'd try. Here's what we want to fund in 2026! 1. Golden Artifacts: Think Chainguard but more broad. Artifact attestation exists for open source. Almost nothing exists for internal software — especially the vibe-coded tooling now running in production. We want the company building cryptographic proof of secure software delivered from secure artifacts: who built it, how, and whether it was reviewed. If more things are being yeeted into the world via Claude Code (myself included), this feels like an issue. 2. MCP & Agentic Security: Agents are getting real credentials and taking real actions. The security posture of most orgs around this is basically zero. That changes fast. You'd never give an employee hardcoded API keys or write access to your email without supervision/trust. Why give it to agents? 3. AI Governance: Boards are asking CISOs to account for AI risk. CISOs have no good answer other than "Palo has a module" 4. Next-Gen Endpoint: CrowdStrike was built for a world of static binaries and human operators. AI workloads, cloud-native infra, and AI-assisted attackers need a new architecture. The category is ready to be reinvented. 5. Networking in the AI Era: Zero trust was designed for humans. What does network security look like when the entity requesting access is an agent? Nobody's really solved this. 6. Email Security + Next-Gen Phishing: LLMs have made spear phishing infinitely scalable. I've never truly understood why Abnormal and KnowBe4 aren't one company. Maybe this time it's different. 7. Frontier Security Lab: We'd back a credible, well-staffed lab focused entirely on red-teaming models and setting the evidentiary standard the industry needs as LLM built apps become the norm. 8. Dependency Security: That Actually Remediates Malicious and vulnerable dependencies are a top attack vector. The tooling is mostly noise — scanners that don't close the loop. The winner here ships fixes, not just alerts. 9. Critical Infrastructure Cyber: Data centers, satellites, power grids, undersea cables. The physical backbone of the internet is increasingly exposed and wildly under-defended. We have data centers in space, for God's sake. Surely we need better cyber for critical infrastructure? 10. PAM for the Modern Era Legacy: PAM was built for static roles, human users, on-prem directories. Cyberark was founded in 1999.....Agents, ephemeral workloads, and cloud-native infra have broken all of those assumptions. Is anyone rebuilding this from scratch? If you're building in any of these areas — or something we haven't thought of — reach out. james@crv.com

450

49 Comments

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

"Build the event you wish existed" That's what we at Monadical did last week at #NYTW. We wanted an AI Engineers discussion for Engineers. There's always a lot of events out there for VCs to network, or for startups to learn about G2M, but very little on lessons learned from actual engineers in the field. So that's the event we hosted. Big thank yous to our awesome panel, Roy Pereira, Ben Cohen and Corey J. Gallon. Here's the key takeaways and the AI tools we're using. 🚀 All three panelists independently called AI Agents the most transformative LLM application they’ve used. They specifically called out Claude 3.5 Sonnet for its accuracy and reliability. 🪨 We dug into how LLMs are “jagged”, not general. They can be shockingly good at some tasks and completely fail at others. Everyone agreed: good evaluations are critical (and hard.) 🧪 Corey noted how public benchmarks and reality are two different things. Most public evals are saturated or gamed. ♊ Ben emphasized that AI projects are actually two projects: building the tool and building the evaluation process. 🧱 We explored how falling dev costs may impact startup defensibility and labor demand. Roy shared that founders are already shifting strategies in response. ⚒️ In a world of daily AI launches, the panel discussed how they decide what’s worth attention, and what’s just noise. They called out tools like Goose, Aider, Claude Code, and Monadical’s own Cubbi, which helps run agentic workflows safely in dev environments. (links in the comments). CTA: What would you want to hear in an AI Applied Engineering talk you attended?

33

5 Comments

After raising $110M, Israel-based NeuroBlade took an unconventional route—its engineering team has joined Amazon’s Annapurna Labs in a “quiet exit.” Rather than a formal acquisition, this talent and tech absorption reflects a growing trend in deeptech: strategic integration over full-scale M&A. For VCs, it raises important questions about how we define “exit success” in frontier hardware startups, where partnerships can deliver as much value as liquidity. A reminder that innovation transfer doesn’t always require the spotlight—it sometimes just needs the right ecosystem to scale. Details: https://lnkd.in/gfGFAvMS #VentureCapital #DeepTech #MergersAndAcquisitions #IsraelTech #Amazon #HardwareStartups

5

1 Comment

𝗪𝗵𝘆 𝘄𝗲 𝗯𝗮𝗰𝗸𝗲𝗱 𝘁𝗵𝗲 "𝗢𝗽𝗲𝗿𝗮𝘁𝗶𝗻𝗴 𝗦𝘆𝘀𝘁𝗲𝗺" 𝗼𝗳 𝘁𝗵𝗲 𝗔𝗜 𝗖𝗹𝗼𝘂𝗱. When we first met the team at hosted·ai, one stat blew us away: AI workloads often consume only 40% of the GPU capacity they are paying for. In the Gold Rush of 2026, everyone is selling shovels. But hosted.ai is building the automated machinery that makes those shovels work 24/7. They announced their Series A to scale their mission of making GPU infrastructure simple, affordable, and—most importantly—profitable for providers outside the "Big Three" hyperscalers. It’s rare to find a team with this much "scar tissue" in infrastructure (25+ years of experience) tackling the most pressing bottleneck in tech today. Onward and upward! 📈 #Startups #Founders #AI #GPUaaS #InvestmentNews z21 Ventures Raj Shekhar Singh Jyotika Gupta Sudarshan Ravi Udit Batish Carl Fritjofsson

15

Voice is a super interesting modality right now - maybe the first modality we're seeing move to open source models across a number of scale ups / enterprises. Reliability concerns, high costs, and open source model performance are pushing engineers to do their own fine tuning vs. relying on third-party vendors of proprietary models. Many of these orgs have already been collecting their own first-party data and now with third-party vendors like Extrian, David AI, etc they can train really high quality models. RL has been insanely hyped, but it's been unclear how long it will take scale ups and enterprises to actually lean in. Voice AI might be hitting that inflection point faster than expected.

45

7 Comments

Honored and excited to partner with Drew Baglino and the incredible team at Heron Power as they re-imagine the electric grid for the modern world. Power in America is not zero sum. We deserve access to cheap and abundant electricity and we should win the AI race. And we must build the capacity to manufacture critical technology here at home. When faced with existential challenges, we deploy technology against them. And we win.

54

1 Comment