😸 YC is watching you

gm legends, happy Friday.

77 builders launched this morning for a shot at YC S26 — Gustaf Alströmer is reviewing the top ones on the special leaderboard; one wants to replace the resume with your GitHub commits; one breaks AI agents on purpose before users do; and forum founders are sharing their wildest launch horror stories.

GitHired was built by Raghav Bansal, a 3x founder with six hackathon wins, and it ranks job applicants by what they actually shipped on GitHub — real repos, real commit complexity, real tech stack — and flags fraudulent contributions before they ever reach a hiring manager.

🔥 Our Take: Hiring engineers has always been an information problem. Resumes tell you what someone claims. GitHub tells you what they built. The fraud detection on commits is the part that matters most — it's not just more data, it's better data. Getting employers onto a new hiring platform is always the hard part, but topping the leaderboard today with 40+ votes suggests the frustration with the old way is real.

Fabraix is an adversarial testing platform from ex-Meta engineers that runs 1,000+ adaptive attack strategies against any AI agent in a blackbox environment — no integration required — to find exactly how it breaks before a real user does.

🔥 Our Take: Everyone building AI agents knows they're fragile. What nobody has is a clean answer for finding the failure modes before launch. The blackbox approach is smart because you don't touch the agent's code, and the attacks adapt to what they find rather than running the same fixed script. This is the kind of QA infrastructure that becomes table stakes in a year. Right now it doesn't really exist.

ElevenLabs has been the go-to for voice for a while. Now they've turned that expertise into agents that actually get things done. You set one up, it talks like a real person, listens, responds, and helps handle the task — support calls, bookings, whatever the job is. Not a demo, not a "press 1 for sales" situation. It's ready to deploy. Feels like one of those shifts where the interface quietly changes. Less typing, less clicking, more just saying what needs to happen and letting it play out.

Flare is a voice-first social app for Gen Z where you capture moments through voice notes, photos, or mood entries and an AI called Orb builds memory and friendship context around them — no likes, no followers, no strangers, no algorithmic feed.

🔥 Our Take: Every generation gets one "we built this because the existing thing is broken" social app. For Gen Z, the existing thing is TikTok and Instagram, where the algorithm is the product and you are the content. Flare's bet is that voice and intimacy win when everyone is exhausted by performance. Whether an AI Orb feels personal or uncanny is going to be the whole question — but that tension is exactly what makes it worth watching.

Imed Radhouani (@imed_radhouani), founder of Rankfender, opened with: "The meeting where everyone nods and no one disagrees. Not because they agree. Because they have checked out." Then listed nine more from there.

The replies built on it rather than argued. Thami Benjelloun from Mailwarm added the revenue-stage version: his team had three different answers for why customers were paying, which meant they were selling confidence, not the product. Maliik identified the solo-founder equivalent: an active git log with a launch date that keeps sliding.

Othman Katim landed the sharpest line: "People stop failing in small ways, so they start failing in big ways."

Good thread if you're building something right now and everything feels fine.