😸 Your coworkers rated you

gm legends, happy Tuesday.

Today: an agent that asks your old coworkers what you were really like, a user interview that watches your face while you answer, and China's food-delivery giant casually open-sourcing a trillion-parameter model. Plus the forum on everyone's new dream employers.

Badge sends AI agents to collect anonymous reviews from people you've actually worked with, verified by work email, and turns them into a portable Trust Score you can put on a resume. Lokesh Motwani built it after 15 years of hiring engineers and reading resumes that told him what people claimed, never how they worked.

🔥 Our Take: An AI just cold-messaged your old coworkers to ask what you're really like. That's Badge, you hand it your contacts, it works out who actually worked with you, asks them anonymously, and boils the answers into a Trust Score for your resume. And fair enough, honestly, resumes are cooked, AI writes them all now and everyone sounds like employee of the decade. The bit that makes it real: you don't pick who gets asked, the agent does, so your five work besties can't carry you. That's a reference, not a favor, and Lokesh spent 15 years hiring engineers, he knows the reference call is the only part of hiring anyone believes. The gross part is you earn points by reviewing others, which is exactly how LinkedIn endorsements became confetti, and everyone starts at 60 like some cursed video game. Slightly dystopian, probably inevitable. I'd still check my score.

Mira runs your user interviews end to end, recruiting, moderating, analyzing, and watches how people feel while they answer, using facial coding, voice emotion, and eye tracking to catch the moments when words and face don't match. It's from Entropik, which has spent nine years and 17 patents reading human behavior for brands like Unilever and Nestlé.

🔥 Our Take: "I'd totally pay for this" is the sentence that's bankrupted more founders than any rival, because people lie in interviews to be nice and then never open their wallet. Mira watches the stuff you can't fake while you say it, your face, your voice, where your eyes dart, so when someone chirps "I love it" with a face full of confusion, it clocks it and pushes. And this isn't three kids who discovered the emotion API last month, Entropik's been doing this nine years with Unilever and Nestlé signing the checks. Two things stop me short. Reading feelings off faces is genuinely disputed science, smart people argue about whether a frown means the same thing in Osaka and Ohio. And something recording your face and voice to judge you is a privacy horror they breezed right past. Creepy, contested, and still more honest than a transcript. I hate that I want to try it.

Framer 3.0 ships AI Agents that live inside your canvas and can actually design. Not suggest. Design. Drop a brief and watch them build out full pages, create components, write code, connect your CMS, and handle SEO. All without leaving the canvas. They plug straight into Claude Code and Cursor. Yes, inside your actual component system, without breaking it. The Figma file you were about to hand off to engineering? Keep it.

LongCat-2.0 is a 1.6-trillion-parameter open-weights model, MIT licensed, with a million-token context window, from Meituan, the Chinese delivery giant. The part turning heads: it was trained on 35 trillion tokens entirely on AI ASIC superpods, no NVIDIA involved, without a single rollback.

🔥 Our Take: The company that delivers dumplings across China just open-sourced a 1.6-trillion-parameter model, and that's not even the story. The story is the silicon: the whole run happened on AI ASICs, 35 trillion tokens without a rollback or a loss spike they couldn't recover, and anyone who's watched a big training run knows that's exactly where alternative hardware usually dies. It didn't. There's an assumption baked into every frontier lab's valuation, that serious models require NVIDIA, and every launch like this chips away at it. For your stack it's open weights and a million-token context you can download today. For the industry, it's a delivery company demonstrating that the hardware monopoly is optional.

Max Musing (@maxmusing), yes, the Basedash founder, asked a simple one: what are your top three dream companies to work for right now? The answers say more about 2026 than any trend report.

Almost every list was AI labs. Anthropic, OpenAI, DeepMind, Cursor, over and over, to the point where Levi (@levi_mitchell1) called it: "three out of four picks in this thread are AI companies now, would've been a totally different list two years ago." Even Max conceded he'd be surprised if anyone in tech doesn't have a big lab in their top three.

The counterpoint came from Nika (@busmark_w_nika), who used to dream of Meta and Netflix and now wants none of it: independence, with the dream companies as clients instead of employers. "It is always heart-warming when employees from a big company subscribe to my newsletter with their company mail."

Good thread if your own top three has quietly become Anthropic, OpenAI, and whoever's paying in GPUs.