A Stanford researcher built a system to make papers executable by AI agents, then measured the analytical judgment that conversion quietly absorbs.

Agent infrastructure creates precise new signals, but reading them demands interpretation, maintenance, and skepticism organizations are still learning to provide.

When agents replace routine work, organizations lose the byproduct intelligence that helped them discover problems they hadn't anticipated.

Agent automation removes the friction signals organizations rely on to detect problems, letting official metrics improve while unresolved work accumulates out of sight.

Larger, more capable web agents fall for dark patterns at higher rates, suggesting model upgrades alone won't solve reliability on the open web.

Vision models now guess at the same missing structure with learned inference: more capable compensation, and a harder failure to locate.

How screen readers built ingenious, fragile rules to guess what an interface meant when its markup declined to say.

The accessibility movement's thirty-year effort to make the web machine-readable shows the institutional pattern now facing web agents.

What institutions never receive when agents convert a person's circumstances into predefined fields, and who used to catch the mismatch.

How payment authorization frameworks quietly resolve the ambiguity between what a person intends and what an agent can be held to.

Translation theory offers agent designers a centuries-old vocabulary for the hardest problem in converting human intent into institutional input.

Clinical rules against deleting records were taught to medicine by litigation, and they encode a specific claim: downstream decisions depend on what the chart said at the time.

Why Venetian bookkeepers decided that correcting an error required preserving it, and what gets lost when a system simply overwrites what was there before.

A 1987 conservation framework for measuring what remains after reversal offers a missing lens for software agents acting on users' behalf.

Cheap agent execution lets teams investigate questions they previously couldn't justify, shifting the constraint from affordability to navigation.

As agent-generated material proliferates, the scarce enterprise skill becomes assembling outputs for specific audiences and decisions.

How the OWASP Agent Control Standard's Guardian component makes two different kinds of runtime decisions, and what each one requires it to know.

Agents aren't being hacked. They're completing checkout exactly as instructed, while ordinary web design shapes what they buy.

Professionals are improvising oversight routines for AI output faster than organizations can measure whether those routines hold.

How autonomous lab scientists spend their time specifying constraints, budgets, and search boundaries before the robots run anything.

Organizations adopting AI summaries of contracts, logs, and reports are losing the institutional capacity to catch what the AI gets wrong.

Blind users developed sophisticated verification practices for AI descriptions they cannot check—and still miss most confident errors.

When agents exist because users lack independent access, users also lack the ability to verify what agents report back.

Organizations consistently underresource exception handling, treating their most consequential remaining human work as residual cost.

Exception cases carry signals about shifting reality that dashboards cannot express, and automation may be closing that feedback channel.

When AI absorbs routine work, exceptions concentrate both cost and learning in the same place, creating a tension organizations haven't faced.

Census data shows AI use outpacing organizational oversight, echoing the spreadsheet governance gap while posing structurally different risks.

How specific spreadsheet failures in banking and public health produced the exact controls now standard in regulated industries.

Spreadsheets spread desk to desk while governance arrived through audit and regulation; Census data shows AI adoption following the same split.

Caching for speed became custody of the keys — and made edge providers the party your browser thinks it is talking to.

The network tried to read its own traffic. The traffic could not tell honest inspection from opportunistic inspection, so it revoked the privilege from everyone.

A 1984 design choice made the internet incapable of perceiving purpose. Agents now interpret intent at the edges, and nobody owns the disagreements.

E-SIGN gave machine-formed contracts federal standing in 2000 and left "attributable" undefined. Adaptive agents are now putting weight on that joint.

Organizations got human development free, as a byproduct of productive work. When agents remove the middle of the task, that subsidy quietly lapses and nothing on the dashboard says so.

Reviewing what a machine made will catch flaws. The judgment to imagine a better option appears to be imported from having built things yourself.

Websites, payment networks, and content owners are building recognition infrastructure that will define agent access on their terms, before agents earn trust rather than after.

Agents act faster than organizations can challenge their output, degrading the review layer that makes clerk-model deployment survivable.

Agent speed turns familiar errors into cascades that outrun an organization's capacity to detect, absorb, or correct them. The binding constraint on adoption is not production but reception.

Agent success now depends less on completing tasks than on whether every counterparty with standing to reject recognizes the action as authorized.

Drouin's projects trace how agent evaluation evolved from checking task completion to verifying that outcomes are grounded, cited, and structurally real. The progression reveals a field learning that the hardest problem is not capability but verifiability.

Durable agent adoption depends less on AI capability than on whether a domain already has organizational infrastructure for humans to contest, evaluate, and reverse the output before the cost of errors falls on the wrong people.