Silicon Valley's AI Colleagues: Petri Dish, Then Brain Fry
Les Echos report (Florian Dèbes) from San Francisco: AI agents already integrated as colleagues at start-ups, "petri dish" (Aaron Levie / Box), reflex use of Claude before every meeting, personal Jarvis, 5 parallel agent tabs, "the limiting factor is human cognition" (Patrick Joubert / Rippletide), "brain fry" / cognitive overheating, BCG/HBR study showing 14% of employees overwhelmed, "token-max" ranking of the heaviest AI users, testimonials from Sinaï/Bangay/Allali/Hodjat/Pantera/Chapeau and an echo from Siddhant Khare ("AI reduces production costs but raises coordination costs").
By Florian Dèbes// Source lesechos.fr ↗/Reading 2 min/.md// Auto-verified translation
#Silicon Valley#San Francisco#AI agents as colleagues#petri dish#Aaron Levie#Box#Florian Dèbes#Les Echos
Summary
On April 22, 2026, Florian Dèbes published a report in Les Echos from San Francisco on the integration of AI agents as full-fledged colleagues at Silicon Valley start-ups. The sentence that structures the entire article comes from Aaron Levie, CEO of Box, quoted by the New York Times: "Silicon Valley right now is a real petri dish."
The report first documents the boost. Justin Bangay (Airbyte salesperson) has Claude prepare every client meeting from previous recordings and the web: "It takes a minute, I save almost half an hour." An investment fund partner has Claude scrape LinkedIn and ZoomInfo before he wakes up to deliver a daily sales brief. Sarah Allali (Lobby) prepares her fundraising through Claude, which lists investors and shared LinkedIn contacts. Logistical emails are delegated to agents — some even sign their replies to make this explicit. But Allali immediately points out the blind spot: "Humans have an ego. Nobody wants to know they're not important enough for someone to bother writing to them."
On the engineering side, Jérémy Chapeau (SubImage) reports having shipped five major features in one week — "Without AI I would have shipped only one". He built his own agent named Jarvis (an Iron Man reference ubiquitous in the Valley) that orchestrates action plans and responds to alerts from another agent monitoring customer behavior. Patrick Joubert (Rippletide) practices maximal delegation, 5 parallel agent tabs, and formulates the central aphorism: "The limiting factor is human cognition."
Then comes the flip side. Babak Hodjat (Cognizant) notes that AI causes fatigue "when you delegate too much, the result is mediocre, and you have to redo everything". The article relays a viral post by Siddhant Khare (Germany): "You use AI to be more productive. So why are you so tired?" His economic thesis: "AI reduces production costs, but raises the costs tied to coordination, verification, and decision-making. And those costs fall entirely on humans." A BCG/Harvard Business Review study (Julie Bedard, March 2026) puts at 14% the share of employees overwhelmed by the pace imposed, with cases of "brain fry" (cognitive overheating, headaches, slowed decision-making). The token-max — an internal ranking that rewards the heaviest AI users — fuels this exhaustion.
The article closes on an anxiety shared by the builders themselves: "Those who adopt it, or even build it, wonder whether they're digging their own grave." Eric Pantera (Swile, Montpellier), however, notes that the SF/Europe divide has largely disappeared for those willing to engage: "The gaps with our friends at Meta aren't significant."
A pivotal 2026 piece on everyday life with AI agents, which simultaneously records the productive success and the first wave of cognitive exhaustion among early adopters.
Key takeaways
Date and author. April 22, 2026, Florian Dèbes, Les Echos ("Vie au travail" section).
Pivotal quote. (Aaron Levie, CEO of Box, in the NYT): "Silicon Valley right now is a real petri dish." A metaphor that structures the entire article: SF as a full-scale experiment.
Temporal framing: "three and a half years after the emergence of ChatGPT" — the article documents the second-wave effect (chatbots → AI agents) at companies that caught the wave early.
Key testimonials (8 named individuals).
Aaron Levie. (CEO of Box): framing NYT quote.
Paul Sinaï. (founder of Blaxel AI): "Here, the ordinary changes overnight."
Justin Bangay. (Airbyte salesperson): Claude drafts the prep sheet for every client meeting. "It takes a minute, I save almost half an hour.""The question is how fast can you go?"
An investment fund partner. (anonymous): Claude reads LinkedIn and ZoomInfo every morning before he wakes up to update a prospect spreadsheet and produce a sales-priorities report.
Sarah Allali. (founder of Lobby, an agent that acts on an email rather than writing it): "Humans have an ego. Nobody wants to know they're not important enough in the eyes of their counterpart for that person to bother writing the message themselves." Claude prepares her fundraising (list of investors + shared LinkedIn contacts).
Babak Hodjat. (AI director at Cognizant): "The future will belong to agents, but we're only at the beginning: the contrasts between start-ups and larger organizations are significant." Later: "AI fatigue can set in when you delegate too much to AI, the result turns out mediocre, and you have to redo everything."
Eric Pantera. (CTO of Swile, based in Montpellier): "The gaps between what our friends at Meta do and what we do in Montpellier aren't significant." Shows that technology diffusion is now near-instantaneous for those willing to engage.
Jérémy Chapeau. (senior engineer at SubImage, cybersecurity): "This week I shipped five major features. Without AI, I would have shipped only one." Built his own agent, "Jarvis" (a common Iron Man reference in the Valley). Practices tight oversight: "By following what the AI agent does as it happens, I can correct it before it drifts too far."
Patrick Joubert. (CEO of Rippletide): "The real value is letting AI code on its own for as long as possible. We review everything at the end, and if it doesn't work, we throw it out."5 parallel agent tabs. Aphorism: "The limiting factor is human cognition."
Concrete practices documented.
Claude before every meeting to produce a prep sheet from previous recordings and the web.
Agents before waking up to scrape LinkedIn/ZoomInfo and produce the day's sales brief.
Email signatures indicating that an agent wrote the reply, for logistical questions ("available at 3pm?").
Token-max. internal ranking that rewards the heaviest AI users — "employer pressure", a source of exhaustion.
Multi-tab 5 simultaneous agents: the standard practice at Rippletide.
"Brain fry" / cognitive overheating.
Widely shared blog post by Siddhant Khare (Indian, based in Germany): "You use AI to be more productive. So why are you so tired?"
Khare's pivotal analytical quote: "AI reduces production costs, but raises the costs tied to coordination, verification, and decision-making. And those costs fall entirely on humans."
BCG / Harvard Business Review study. (Julie Bedard, early March 2026): 14% of respondents say they use AI without feeling able to keep up with its pace. First cases of "brain fry": headaches, slower decision-making.
Khare's cited advice: don't use AI if the result isn't 70% correct after 3 attempts, preserve AI-free time to think.
Structural tension of the article."Those who adopt it, or even build it, wonder whether they're digging their own grave." Record number of open software positions in the US in Q1 — optimists see confirmation in this, pessimists see the last push before total automation.
Adoption status. Claude Code and Codex had limited access a year ago. Rapid diffusion outside SF/SJ but variable depending on the organization.
Connection to the watch dossier.
Confirms the Mollick / Bersin narrative on shadow adoption, but observed here out in the open.
Provides a French empirical basis for the MIT NANDA finding: the technology works (the boost is real among early adopters), but introduces new cognitive costs.
Cross-references existing Rippletide notes (Patrick Joubert, already covered in October 2025) and adds complementary "everyday life" evidence to the theoretical harness/agent framework.
The token-max concept is an interesting signal to watch — an AI version of a "shadow performance review."
Key figures
14% des salariés dépassés par le rythme imposé par l'IA