All writing
What taking control requires
A cognitive ergonomics review of Hermes Bot Screen: five findings about human control, observation and state, reproduced and corrected before merge.
You move the real work
How AI affects working conditions: workload, exceptions, judgment, meaning. What an AI project moves, and three questions to ask before automating.
Agents are operators
Benchmarks test models. Real work tests situated operators: model, harness, tools, permissions, evidence, recovery, and human belief.
Making Grok act
An XML system block that forbids intent narration: one repro (10 min → 33.9 s), corpus measures, an honest A/B, and what the patch cannot fix.
The problem is belief
An agent is not enterprise-ready because it can act. It is enterprise-ready when the belief it creates about its action is calibrated to evidence.
AI agents in SMEs: real work first
A useful AI agent starts with a situated task, limited rights, verifiable evidence, and a possible human recovery path. In Réunion and elsewhere.
Talk about real work
Why a useful AI talk should start from what AI changes in work, decisions, responsibility, evidence and human recovery, not from tools alone.
Raw traces are not evals
The missing layer between real agent failures and measurable model progress: reducing messy traces into replayable eval seeds without laundering the signal.
Tool use is not completion
Why agent reliability depends on preserving the boundary between intention, action, observable state, and truthful final claims.
Claims without evidence
Why the critical risk is not only hallucination, but an agent claiming work is done without observable evidence.
Agents are work systems
Why task completion is not enough: an agent redistributes verification, responsibility, coordination, and recovery work.
AI reconfigures work
Why AI must be analyzed from real activity: constraints, trade-offs, invisible cooperation, room to act, and responsibility.
The benchmark lie
A cognitive ergonomist dogfoods Grok 4.20 across 12+ production agents. What 40 years of human factors research says about what benchmarks miss in agent loops.
What the AI tool displaces
Opening note. What AI concretely displaces in the ergonomist’s practice and in organizational work — and the analysis space it opens.
Case studies
A consulting firm: an agent for producing deliverables
Consulting firm
An engineering team: making AI a shared matter
Engineering firm
xAI: when the agent says “done” and nothing was done
xAI
Nous Research: my code merged fifteen times into Hermes Agent, one review of Bot Screen
Nous Research
AMOI: the agent takes the paperwork, the metalworker stays on site
AMOI