Meet @IgorBlinkk , 17 years young, who came from a village in Kazakhstan with less than 2,000 people. Despite having internet speeds of ~3 Mb/s, he built multiple apps including a game that grew to 12M+ users. When our investor found him, he traveled 2 hrs to the nearest cafe for the wifi to take the call. He is the first person in his family to leave Kazakhstan, now in SF building a generational computer use agent data company. This is what the FIR U cohort at Afore Capital looks like.

we're running it back w/ robots, researchers, engineers & founders interested in physical AI @AforeVC next Tue in SF (oh and sushi + pizza) - RSVP to get on the list if you want to catchup w friends @sundayrobotics @physical_int @AppliedInt @BedrockRobotics @SkildAI @Tesla @GoogleDeepMind and more! partiful.com/e/aiB92GgPlsLbF…

For 10 years, @AforeVC has invested in People Before Ideas. Today our new website puts that front & center 🚀 afore.vc/residence Here's what #PeopleBeforeIdeas means to us 👇

@AforeVC giving us a little preview of life after Series B Rental suite at Oracle tomorrow for Angels @ Giants Neither team is good. Seats still slap. I bet YC doesn’t do this Upcoming FIR batch is coming up. Happy to refer you if you’re cracked

Gemini 3.6 Flash (the 3.5 Flash upgrade) is SOTA on MMMU-Pro, the benchmark for multimodal reasoning and VQA. We just built a benchmark it can’t pass. Real visual questions from people across 100 countries. Counting, spatial reasoning, OCR, text in the wild. Every question here has already stumped a frontier model. We did this in a week. The current SOTA gets 41% of the way there on v1 of the benchmark. The breakdown is rougher than the headline. Naming what it sees: 88%. Counting what it sees: 33%. Five tries per question, and 17 tasks failed all five. Models become as good as their benchmarks. Today’s benchmarks are full of things models already do well, like coding. Real-world perception is not one of them. Karpathy says AGI is a decade away. This data says he might be right. AI labs: if you want your next model to actually see the real world, hmu. Everyone else: what’s the simplest image that still breaks every model? 👇

Most people today do their research and get recommendations from ChatGPT, Gemini or Claude for almost everything. Commercial signals that used to flow through search engines now flow through LLMs. SEO & SEM is now obsolete. @Petra_Labs has a shot at replacing them 🚀

FIR founders come from everywhere, and they show up already deep in the work. Take the team at Percept. Alper Gel has been doing computer vision research since his freshman year of high school, nearly a decade in. Victor Wu was doing computer vision research at Mayo Clinic before Percept. Samuel Cornelison helped build two public safety startups that were both acquired. Now, at FIR, they're building the spatial memory layer for robots in large, unstructured environments. Want to join the next cohort? Link is in the comments.

We just wrapped the first week of FIR S26. Rather than sum it up ourselves, we'll let a couple of the founders do it: "I've had an amazing first week at FIR. Being surrounded by ambitious founders and having great conversations with them has been invigorating. Keen to build and excited for what's to come." - Joseph Dowdall "It's great to be in the Afore offices with so many brilliant founders, investors, and advisors. The first couple weeks have been full of ideas, best practices, and hearing from people who have been in our exact situation before." - Victor Wu They’re not just building: sushi & soccer dinner at our office, a workout class, and a farm-to-founder dinner at Plow to close out the week. 21 founders, 12 teams, one summer just getting started. More to come, and if you want to be part of the next cohort, applications are open. Link in comments.

FIR founders don't take the normal route. Take the team behind Verso. Jonathan Bernal came to the US, taught himself the language, trained as an engineer, and worked on AI at one of the biggest companies in tech. Drew Carson studied computer science, dropped out to chase startups, and built and sold a consumer company before he was done. Two very different paths that ended up in the same place. Now they're building together in the FIR cohort. Two founders who've done hard things before, doing it again. That's what FIR looks like. If you're the kind of person who takes the hard road to build something that matters, the next batch is open. Link is in the comments.

Today we’re giving you the first real choice in where to run your .xlsx files. We’re also announcing our pre-seed funding, led by Afore and joined by many prominent AI founders who share that conviction. Excel is one of the greatest tools I’ve ever used. I grew up on it. I was forged by it. It is one of the great inventions of humankind. But for forty years, the world’s most important file has had only one trusted way to run it: Microsoft Excel. Today, there are two. Introducing Nobie. It will soon reach 100% feature parity with Excel. We’re building it for the 1 billion humans using Excel today and the 10 billion agents that will work with spreadsheets tomorrow. Nobie is now in beta, available as a native Mac app and as a CLI on macOS and Linux. It’s free, and everything you can do in Excel will always be free in Nobie. Nobie isn’t done. But the product you’re looking at today is a glimpse into the near future: what you and your friends are going to reach for when working with spreadsheets. We are dedicated to a singular purpose: delete every reason you still need Excel. Before long, it will seem absurd that Excel was ever all we had. Claude Code and Codex are built right into Nobie. Or bring your favorite agent, like Antigravity or Pi. Robots of all kinds are welcome. So if you’re a human on a Mac or an agent on Linux, try it out and let us know what you think. We think you’ll soon make Nobie the default app for .xlsx. Enjoy!

High school friends from Milan and LA, now building Krypton AI together in San Francisco. That's Ludovico Cordara & Ray Cabrera, two of the student founders who joined FIR U this summer. What they say about the program says it best: being surrounded by talented student founders working into the early morning pushes you to keep going. And while we're proud to be their mentors and investors, the best part is that every conversation feels like one between equals! If you're a student founder with that kind of drive, we want to meet you. Applications for the next FIR U cohort are open. Link in comments.

FIR U founders are building things most people think are still decades away. Take Subvocal - Arnav & Aarush. They're building a wearable that reads your inner voice, the words you say in your head before you ever speak them. They've already hit 45% accuracy on any word in the English language, trained on less than two hours of data. Two hours. And it's getting sharper every week. This is what the FIR U cohort looks like. The next batch is open. If you want to build something this ambitious, link is in the comments. Apply now!

never forget your first believer when @AforeVC invested in me, @hyperspell didn’t exist i was in the middle of a cofounder breakup in my last company and starting over from scratch i hadn’t met my cofounder @maebert yet there wasn’t even a company to invest in they weren’t betting on traction they were betting on a person today we’re serving fortune 100 customers through every stage, @AforeVC has been so much more than an investor and has never stopped believing in us they’ve helped us win customers, hire great people, and become real friends and mentors along the way i’ll never forget that first bet

Founders love to say they're "cooking." So we put it to the test with a California cooking class. Safe to say, they're cooking!

Congratulations to Hightouch for landing on Redpoint's InfraRed 100, alongside Anthropic, OpenAI, Databricks, and Lovable. @kashgupta_ , @tejasmanohar , and Josh Curl, well deserved. Proud to have believed in you from day zero 🤝 See the full list and report here: redpoint.com/infrared/100/

Today we’re introducing @Seda_AI_ - The world’s first end to end Synthetic UX Testing Agent 🤖