Data labeling firm predicts enterprise evals will become core IP as revenue shifts to Fortune 1000
soumitrashukla9 · x · 2026-09-26
Aaron Levie argues you can't automate what you can't measure: evals are a gate to enterprise AI adoption, since companies have no way to evaluate non-deterministic agent workflows without them.
A quote-tweet adds insights from an interview with a major data labeling company:
- They expect most revenue to come from Fortune 1000 enterprises rather than labs within a few years
- Every company will want to own its intelligence, but that doesn't require open source models
- Evals will become a company's main proprietary IP, given how much agent performance improves after properly setting up internal eval environments
- Most enterprises haven't moved past coding agents, largely due to lacking eval infrastructure for non-engineering agents
More from coding & agent
- Vercel launches stealth model Pixel Canary, ties GPT-6 Astra on Next.js evals — cramforce · 2026-09-26
- steipete pushes back on agent cost concerns: capable models now $0.1 per million tokens — steipete · 2026-09-26
- steipete: sync SQLite access was my biggest design mistake; 575 PRs to make it async — steipete · 2026-09-26
- Oh My Pi adds custom model backends: vLLM, llama.cpp, SGLang and more — bolts98 · 2026-09-26
- Fighting architecture drift in long ChatGPT projects with an architecture anchor — player_unkn0wn · 2026-09-26
- Claude Opus 5.5 generates a full launch video — animation, music, voiceover — in 20 minutes — cedric_chee · 2026-09-26