Diogo Almeida

launched a model

accidentally served trillions of tokens a day

29.4% of the fortune 500 showed up

raised a really big series A from @a16z

it's been 3 weeks

we would like to sleep now

the gang's all here

building the primitives so builders can cook

Excited to release Step-Jev 🔥

step-by-step rewards for agent RL, even when there’s no answer key

After DeepSeek R1, everyone said step rewards get hacked, so just reward the final answer.

We found the opposite. With an AI judge grading only the final answer, our agent learned to fool it and accuracy dropped every single update until it was worse than no training at all.

Step rewards fixed it. Everything is open!

Excited to release Step-Jev 🔥

step-by-step rewards for agent RL, even when there’s no answer key

After DeepSeek R1, everyone said step rewards get hacked, so just reward the final answer.

We found the opposite. With an AI judge grading only the final answer, our agent learned to fool it and accuracy dropped every single update until it was worse than no training at all.

Step rewards fixed it. Everything is open!

I wonder why Perplexity decided to get into decision models

NEW: Top SaaS Vendors on Ramp (Oct 2026)

We clearly see AI spend pulled in two different directions.

On the downside, new and cheaper models are pushing AI spend down. Jev captured an entire percentage point of market adoption within a month (unheard of growth). And routers continue to gain.

On the upside, the increased adoption of expensive image and video generation models will push AI spend up, like Higgsfield and Runway.

NEW: Top SaaS Vendors on Ramp (Oct 2026)

We clearly see AI spend pulled in two different directions.

On the downside, new and cheaper models are pushing AI spend down. Jev captured an entire percentage point of market adoption within a month (unheard of growth). And routers continue to gain.

On the upside, the increased adoption of expensive image and video generation models will push AI spend up, like Higgsfield and Runway.

Jev launched on a Thursday. By Sunday, the @typesafeai team had served a trillion tokens with no signs of slowing down.

Learn how they scaled Jev on Modal.

Can a model that doesn’t generate text compete with frontier LLMs?

We independently tested TypeSafe’s Jev against 11 other models. On 400 claim-verification questions, it matched GPT-6 Astra’s 97.5% accuracy at ~1/500th the cost.

New: Decision Model Rankings!

See the spend share and token share of different decision models and task types, including Relevance, Correctness, Instruction Following and more.

@typesafeai is leading all categories today.

I benchmarked "Jev-killer" OpenAI Decisions API against Jev for HiringCafe.com, serving 2.5 million users. The task is to predict how relevant a user query/resume is to a job description, on a scale of 1-10. OpenAI is 2x more expensive and 5-10% worse. 🧵

Jev beats Luna Decisions on *11 different MCQ* benchmarks. It may be a decisions API, but it's not nearly as smart as Jev, and it's also twice the cost. Paretomogged, as most jev clones are.

There's something wet-sponge-feeling about the qwen fine tunes that I can't explain compared to Jev

Wrote a short blog about the "shape" of language models, and the tradeoffs they may present in the future. I genuinely think it's a valuable research direction to start thinking about now, especially w.r.t. harness design.

alexzhang13.github.io/blog/2026/shap…

Yes, we are still talking about Jev. Here's my contribution: x.com/i/article/2104…

Jev is now in the AI SDK for Python.

To test it, we ran two experiments:
▪︎ Detecting Python vs. English as text is typed
▪︎ Writing Python, one decision at a time

𝚞𝚟 𝚊𝚍𝚍 𝚊𝚒

vercel.com/blog/jev-for-p…