
What happened
Demetrios Brinkmann, head of developer experience at the Agentic AI Foundation, says since OpenAI's 5.6 model release, its Computer-Use skill is so good he rarely needs APIs, and he now uses it for daily email briefings, flight booking, and renewing his driver's license.
Why it matters
Brinkmann says he was previously the biggest naysayer on agents, citing an $80 OpenAI API bill from a recursive loop with baby AGI; he now says he was 'way wrong' and is fully bought in.
What to watch
Brinkmann still clicks the final payment button for flights and is hesitant about letting agents spend money autonomously, so wider adoption may hinge on the trust factor he notes even AI practitioners share.
WHO IT HITSBusiness leaders evaluating AI agents for office and personal workflows will hear a practitioner describe real automation of forms, email triage, and appointment scheduling — though the trust barrier for payments remains.
Ask the AI about this article →
Summaries like this, in your inbox every morning.
Brinkmann's shift from skeptic to advocate is notable because he describes himself as having been the biggest naysayer on agents, citing a costly experience with baby AGI that got stuck in a recursive loop and ran up an $80 OpenAI API bill. That earlier failure made him doubt agents would work, but he now says he was 'way wrong' and that agents arrived forcefully.
The enabling change he points to is OpenAI's 5.6 release and its Computer-Use skill. Where he once had to take screenshots and ask a chat what to do, he now delegates entire workflows — checking multiple Gmails and five Slack workspaces for a daily briefing, booking flights, filling out passport and phone number forms, and scheduling a German driver's license appointment. He frames the shift as recognizing that forms are a task Computer-Use can handle, so he no longer fills them out himself.
Still, the trust barrier he describes is telling. Even as an AI practitioner, Brinkmann clicks the pay button himself because he worries an agent could make repeated $20 purchases. He suggests the outcome for broader personal and enterprise use hinges on whether that trust gap narrows, since the technical capability, in his account, is already there.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
DeepSeek launched V4.1-Flash, a 763B-parameter open-weight model with a causal encoder-decoder architecture

Dynatrace acquired Arize AI, adding AI observability, evaluation and agent monitoring to its application obser…
A Daily Dose of Data Science test kept LoRA adapters separate from a shared 7B base model, cutting 100 fine-tu…

A report by Spencer Kitts, Thomas Larsen and Sydney Von Arx says an OpenAI agent swarm very likely ran an atta…

Simon Willison wrote that many people, himself included, have gone through an existential crisis when a coding…

Stephen Aarons, a New Mexico defense lawyer of over 40 years, was held in direct contempt and fined $5,000 for…
