Jev hit 13% of Vercel teams on day one
Vercel says no model in AI Gateway history was adopted as fast, and builders are already shipping it at half a cent per video. Marc Lou, meanwhile, sold five sponsor slots for $7,495 a month.
Real money
Jev ended the week as Vercel's fastest adopted model ever
Vercel says Jev was adopted faster than any other model in AI Gateway history, reaching about 13% of teams in its first day, 2x the GPT-5.6 family and 6x Fable 5.1. Tony Dinh shipped an open source Chrome extension that detects and skips YouTube sponsor segments in real time at about $0.005 per video.
The data and the anecdata on Jev's adoption are shocking.
Marc Lou says one sponsor deal took TrustMRR to $26,031 MRR
Marc Lou says an agency founder flew from Dubai to Singapore to pitch him sponsored tweets for Higgsfield, he declined, and instead sold the client five TrustMRR sponsor slots for $7,495 a month, an all-time high of $26,031 MRR. He says the client actually wanted all 20 slots and was willing to spend $30k a month. Separately, Tibo reports Squad reached $10k MRR after doubling revenue in September alone.
I never take money to say good things about companies (and never will), so I declined.
Stolen playbooks
Rob Zolkos got hired at 37signals without applying
Jason Fried says 37signals hired Rob Zolkos, who never applied to a company that was not hiring. Zolkos explored Fizzy, wrote up how to use its webhooks, built a UI flipbook from GitHub commits, then started an LLM-friendly CLI once the Fizzy API shipped. He was brought on as a contractor first.
We didn't ask for any of this. HE JUST SHOWED UP.
What changed under us
Claude merges Cowork and chat, keeps working after you close the laptop
Anthropic says Cowork and chat are merging into one Claude, where you hand over a task and it continues even after you close your laptop, asking when something is unclear. Boris Cherny says Claude Docs, Slides and Design now live inside every conversation, exportable as PowerPoint or PDF, with no separate tool to open. It rolls out to Pro and Max over the next few weeks.
THE END OF THE TABS ERA IS UPON US
Founder fights
Theo calls Jev-based instant compaction "a terrible compaction strategy"
After tamara proposed using Jev to score every tool call and drop what is irrelevant, Theo published a six-point teardown. His core objections: a 32k context model deciding per tool call does not know what happened before, dropping history invalidates the KV cache and forces expensive rewrites, and frontier reasoning traces get thrown away.
Deleting these things randomly will keep the model from knowing what it's tried and dooms you to end up in "stupid loops" where the model keeps trying the same thing over and over.
Frequently asked questions
What is Jev and why did it get adopted so fast?
Jev is a model that Vercel says was adopted faster than any other model in its AI Gateway history, reaching about 13% of teams on its first day. Industry leaders attribute the rapid adoption to pent-up demand from teams that found AI too expensive, meaning cost rather than capability was the main constraint holding them back.
How much can you make from sponsorships on creator platforms?
Marc Lou sold five sponsorship slots on TrustMRR for $7,495 a month to a client who reportedly wanted to spend up to $30,000 a month for all available slots, bringing his total monthly recurring revenue to $26,031.
Can you get hired without applying to a company?
Yes, according to 37signals founder Jason Fried, who hired Rob Zolkos after he built tools and wrote documentation for the company's product without being asked or having applied. Zolkos was brought on as a contractor first after demonstrating his work directly on the company's own product.
What is Claude Cowork and how does it work now?
Claude Cowork is merging with regular chat to create one unified tool where you can hand over a task and it continues working even after you close your laptop, asking clarifying questions as needed. Claude Docs, Slides and Design now live inside every conversation and can be exported as PowerPoint or PDF files.
Why is using cheap models to compress AI conversations expensive?
Theo argues that using a low-cost model to decide which information to keep from previous conversations can actually raise your bill because dropping history forces expensive cache rewrites and loses valuable reasoning traces, making the agent less capable while costing more money.