Specs are the new code review
When agents write the code, humans add value in the spec and the evals. How Even's process runs: scope, build in parallel, review agent, evals, human gate.
Insights
Notes from building with agent teams and from working with teams. Short, concrete, opinionated.
What others do
Private platforms vs. agent-native products
JPMorgan, Imperial, Klaviyo, Figma, Chinasoft. From a safe chat for employees to agents wired into the systems. What the DACH mid-market should copy and skip.
The briefing is the product
One day, twelve people, a live demo on real code or a real workflow, a decision at the end. Why that beats a 40-page strategy deck, and how the day runs.
What others do
Legal and documents: 15 hours to 15 minutes
Spellbook, EvenUp, Kimi, Morgan Stanley. Long-context document work is the quietest big win in enterprise AI. Where it applies outside legal.
What others do
What support teams actually shipped
Klarna, Intercom, Autodesk, Starbucks. Read for the mechanism, not the headline: an operator wired into backend APIs, with evals. What a COO should measure.
AI news
AI weekly: cheaper models, harness evals, week of 2026-10-09
Haiku 5.5 moves the small-model price floor, OpenAI's Cookbook shows why tool tests lie, a US open model arrives. What it means for a DACH engineering leader.
Three things a CTO should measure before adopting agentic engineering
Evals, review load and cycle time. If you do not have a baseline for these three before the agents arrive, you will not know whether they helped.
An AI operator is not an assistant
An assistant drafts; a human sends. An operator owns the outcome and escalates when it must. The difference decides whether AI removes work or adds a review step.
Why we build our own products before teaching anyone
Most AI advice comes from people who have not shipped with it. We build Even with agent teams first, then teach. Here is why that order is non-negotiable.
AI news
AI weekly: price steps, stable harnesses, week of 2026-10-02
GPT-6.1 Sol at a fifth of Astra's price, Sonnet 5.5 up to 30% cheaper, Pi 1.0 stable, sandboxes per cloud. What a DACH engineering leader does with it.