Harmonic

A Very Harmonic Officewarming was had. Thank you to everyone in our community who made it so special. And a shoutout to our very own (Nach) for bringing vinyl DJ back through the wee hours of the night x

These angel investors are on a divine heater. Last week we posted the fourth and final Harmonic Hot 25 of 2026, which closes out this year's vintage. A perfect occasion to front-run some of the upcoming year-in-reviews... Here are the angels on the cap tables of the most Hot 25 companies this year. The top 3: Jeff Dean, ex-Google Chief Scientist, co-founder of Discovery Loop (6) Guillermo Rauch, founder of Vercel (5) Fei-Fei Li, Stanford professor, co-founder of World Labs (4) The three split nicely across the AI stack. Jeff's list is power, inference and training infrastructure. Guillermo is closer to tools and apps while Fei-Fei sits in between. The next 6, tied at 3 companies each: Akshay Kothari, COO of Notion Amjad Masad, co-founder of Replit Andrej Karpathy, OpenAI founding member, now at Anthropic Aravind Srinivas, co-founder of Perplexity Karim Atiyeh, co-founder of Ramp Nat Friedman, ex-GitHub CEO, now AI Products at Meta The overlap is real. ResolveAI has 4 of these 9 angels on its cap table, and Wafer, Simile, Julius AI and Prime Intellect each have 3.

✨ New Harmonic Hot 25 ✨ ResolveAI goes back to back to back for the top spot, one-upping Lovable and LangChain as the only startup to top the list three times. And they’re not showing any signs of slowing down, recently adding Jason Forget and Jim Welch from Cockroach Labs to lead sales. Latent is the highest ranking new entrant reaching 2nd place after a $80M Series A led by Spark Capital and Transformation Capital. Starcloud rounds out the top 3. Back on the list, 5 quarters after debuting with 30 people on staff and $200M raised. Data centers in space are looking ready for takeoff. Thousands of investors use Harmonic every day. These are the startups they’re watching most closely heading into Q4: 🥇 ResolveAI: Spiros Xanthos, Mayank Agarwal 🥈 Latent: Sri Somasundaram, Rish Jain 🥉 Starcloud: Philip Johnston, Ezra Feilden, Adi Oltean 4️⃣ Prime Intellect: Vincent Weisser, Johannes Hagemann 5️⃣ Strala: Timon Gregg, Armando Schmid 6️⃣ Emerald AI: Varun Sivaram 7️⃣ Trajectory: Arjun Karanam, Michael Elabd, Ronak Malde 8️⃣ FurtherAI: Aman Gour, Sashank Gondala 9️⃣ Didero: Tim Spencer, Lorenz Pallhuber, Tom Petit 🔟 Periodic Labs: Liam Fedus, Ekin Dogus Cubuk Full list in comments ⬇️

Logan and Asa at Elevate Ventures use Scout via MCP to rapidly evaluate deals (like Potter) against thorough public, private, and exit comps, ultimately driving IC decisions. We walk through their wizardry in Vol. 7 of the Substack: https://lnkd.in/gMzfkr7x

Smoking… survives? You're 4x more likely to raise a Series A as a Juul Labs alumni than an ex-Googler. Ok, maybe not quite that simple as almost 100x more founders have come out of Google since 2020. But it's nevertheless interesting that 20% of founders who last worked at Juul made it to Series A. For founders most recently at Google, the rate is just under 5%. This platform feels full of mafias and founder factories but rarely do we get them quality adjusted. So I pulled the companies who've minted at least 20 founders since 2020 with Harmonic and ranked them by the percentage that reached a Series A. Oculus VR tops the list for companies with at least 20 alumni founders with a 27% Series A success rate. Anduril boasts 19% on 52 founders. Other examples of impressive success rates on high volume are SpaceX with 16% and Cruise at 17% on 137 and 79 founders respectively. Reading through the results, two archetypes stand out. Unsurprisingly, the cohort of respected deep tech brands has attracted software investors in transition. But perhaps more interestingly, notorious startups that hit a public rough patch like Juul, Convoy, Sonder and Cruise, seem to harden would-be founders in ways a smooth ride never could?

Sequoia, a16z and ICONIQ just doubled down on Rillet. Accel and Ribbit backed Campfire. Lightspeed and Khosla backed DualEntry. Seven of the best firms in venture agree on AI-native ERP but disagree on the winner. Our friends at Rillet raised a $100M Series C led by ICONIQ last week. Three years ago this category was one YC check into Campfire. Now it has raised roughly $414M, 93% of it in the last 15 months. NetSuite is the incumbent everyone mentions and at first glance they seem to be doing fine. Revenue is growing 14% a year across more than 43,000 customers and Harmonic's agent, Scout, pegs the combined market share of the three challengers at somewhere around 1% of NetSuite's. But Scout also finds that virtually no startup or scaling SaaS business chooses NetSuite for a new deployment.

"Unless you're constantly refreshing your connections, your networks become stale, and you're less valuable to your portfolio." -Nihar Bobba, Partner at Better Tomorrow Ventures BTV runs on a five-person team, and most of their sourcing comes through people the team knows collectively. So the network has to keep expanding, and all of it has to be reachable by the entire team. 🔍 Scout handles the expansion. Nihar describes the profiles he's looking for and the agent surfaces the best ones. 🗺️ Network Mapping handles the upkeep. BTV synced the team's LinkedIn and email, and Harmonic maps those relationships onto the complete startup database, highlighting warm paths into any company. Through the MCP, Nihar puts all of it to work from inside Claude: "We get portfolio updates every month, and one of them might ask for customer intros. I can port that email into Claude and ask who I know at the relevant companies." The same network helps with the portfolio's hardest hires. Founders ask for narrow profiles, like a fintech engineer who built end-to-end systems through fast growth and Harmonic returns everyone who fits. "Now the whole field's in front of me, and the job is choosing between them, instead of wondering who I missed." -Yoni Lateiner, Head of Talent How BTV maintains its network edge with Harmonic: https://lnkd.in/gVzqAbDc

2 years ago our company data lived in 8 Excel sheets and multiple inboxes. Today it's in 1 MCP server that we call Draper Brain. It connects 15,000+ contacts, 450+ portfolio companies, 4,900+ co-investors, and 1,900+ meeting notes. My whole investment team uses it and so do I. What’s even more exciting is that our MBA interns built it. In their interviews, every single one of them pitched a tool the way a founder pitches a company. We treat them like founders! We gave them the infrastructure, the resources, the access, and got out of the way. What they shipped was remarkable. We now have Draper Brain that pulls from our internal records plus Harmonic's live data on 35M+ companies and answers any question about any company in seconds. The Draper Deal Engine scores incoming deals across eight dimensions, including founder credibility and contrarian thesis. It��s trained on four decades of my own investment decisions. Harmonic surfaces a couple hundred companies for the Deal Engine, and the model takes the first pass. The first week we ran it live, two startups cleared the bar. Incredible! Bet on junior talent the same way you bet on founders. Thank you to Harmonic for telling the story... link in comments.

"When I joined as an intern, our company data lived in eight Excel sheets, and people were dropping them into AI to try to surface information." Michelle Kwok is now a Principal at Draper Associates, which backed Tesla when EVs were niche and Coinbase when crypto was a fad. She manages the firm's AI builds. Draper hands its interns open problems and the tools to solve them. Each one pitches a solution, the way a founder pitches a company. And they've been building with Harmonic. Patrick Blumenthal built Draper Brain, an MCP server that brings Draper's internal records and Harmonic's market data together inside Claude. "Mapping the competitive landscape used to take half a day of research. With Harmonic powering Draper Brain, it takes minutes," -Patrick Jiatao Fan and Spandan Pattanayak built the Draper Deal Engine, which scores companies against four decades of decisions. Out of the tens of thousands of companies Draper could plausibly talk to, it surfaces the few that fit. "People want the frontier, weekly. They don't want two hundred a week," -Michelle Then every company the model flags goes to Scout for a second opinion. "Scout is really ChatGPT plus PitchBook. Whenever the Deal Engine flags a company, I check it against Scout myself, to confirm the model got them right." -Jiatao More on Scout via MCP soon 👀 Read about how Draper's interns built its AI stack on Harmonic: https://lnkd.in/gU44JvGX