Cloudflare says one Workers AI model returned fragments from other customers’ requests. The incident turns tenant isolation and leak attribution into deployment tests for long-context inference.
OpenAI’s activity-level analysis shows broader work use across countries, but its individual-plan sample and omitted categories make it unsuitable as a direct headcount or enterprise-adoption measure.
Cedefop’s review of Estonia’s OSKA system identifies six governance priorities; the practical unit is a traceable path from evidence to an owned policy action.
Google DeepMind’s Genie 4 preview adds native audio and more realistic interactive scenes, while its stated drift and consistency limits define the acceptance tests that matter.
Skills intelligence connects a shared language of skills with evidence about roles, demand, learning order, and change. It goes beyond listing capabilities: the useful output is a decision about what to learn, hire for, build, or stop prioritizing.
This site applies that idea to AI and technical work as a public research project. It is not a commercial workforce platform and does not profile employees.
Read the practical guide →One evidence system, read from four different angles.
467 GenAI and data-science skills explained through practical workflows, examples and primary sources, placed in a prerequisite graph — what to learn, and in what order.
Explore illustrative AI exposure through representative tasks, explicit assumptions and sensitivity ranges. The same work families organize the role dictionary.
423 terms that shaped the AI conversation, 2023–2026 — who coined them, which became standards and which stayed hype. 56% of the discourse is still froth.
A periodic assessment of which skills are rising, holding, or fading — the atlas read against the frontier, published as a point-in-time judgment.
Compare skills taxonomies and ontologies, understand what ESCO, O*NET, and SFIA each measure, and crosswalk public frameworks into a working skills graph.
What RLHF means in modern AI, how DAPO, GRPO, and DPO reshape post-training, and when to use LoRA/QLoRA PEFT vs online reinforcement learning.
Evidence stays attached. Each skill pairs an explanation of the work with a concrete example, known limitations, and primary references. The graph adds a suggested learning order.
Read the methodology ↗
