ClickUp Swaps 22% of Staff for AI Agents as Box's CEO Calls the Cuts Clueless

01A startup hit $300M selling AI budget cuts the same week GitHub started metering Copilot by the token

GitHub Copilot dropped its flat monthly fee and switched to token-based billing. Developers reacted fast. "What a joke," read one comment TechCrunch collected, where the recurring complaint was the same: a predictable monthly line item turned into a meter that moves with every prompt.

The change closes what TechCrunch called the golden age of Microsoft's Copilot, the stretch when a fixed subscription bought effectively unlimited use. Token pricing moves the risk onto the buyer. A heavy coding week now costs more than a light one, and a developer cannot forecast the bill before opening the editor.

That uncertainty is what one heavy user sat down to total up. In a post titled "The solution might be cancelling my AI subscription," the author tallied what the tooling had actually produced: a stack of half-finished projects, a news outlet he says he runs by accident, scripts that never solved the original problem. His reasoning was financial as much as anything. He wrote that he could no longer justify the time, attention, or token spend, and resolved to cancel.

One tier up, the same arithmetic is making money. Glean, an enterprise AI search startup, crossed $300M in top-line revenue and tripled its annual figure even as larger tech companies entered the category. Its sell is no longer only better search across company data. Glean now pitches itself partly on shrinking a customer's overall AI bill by consolidating the separate tools an organization already pays for.

Across the individual, the developer, and the enterprise, the buyer started running the numbers at the same moment. AI is moving from a fixed cost nobody questioned to a variable one finance itemizes. Glean's revenue is the loudest evidence: when "spend less on AI" becomes a feature, the savings have become the product worth selling.

What none of the three resolves is where the new ceiling sits. Copilot has not published the per-token rates that will decide whether the average developer pays more or less than the old flat tier, and that figure is what each team's renewal call now turns on.

02Erin Brockovich Has Spent Decades Forcing Companies to Open Their Files. Her Next Target Is the Data Center.

Erin Brockovich, the environmental activist whose name became shorthand for taking on corporate stonewalling, has picked a new fight. Her target this time is data center secrecy, according to TechCrunch.

The report says Brockovich has a new mission. It does not specify which records she wants pried open. The framing is narrow and deliberate: the issue she names is opacity itself, the way the facilities powering AI keep their operations out of public view. She has built a career on the premise that what companies decline to disclose tends to matter most. Now she is pointing that instinct at an industry that barely existed when she started.

The scale of what she is up against came into focus a day earlier. SoftBank said it will invest up to €75 billion to build data centers in France. The stated goal, the firm said, is to develop and operate up to 5 gigawatts of additional capacity. That is a single company, in a single country, committing a sum larger than the annual budgets of most government agencies to one buildout.

Five gigawatts is roughly the output of five large nuclear reactors. SoftBank framed the spending as additional capacity, language that says nothing about siting, resource draw, or local impact. That gap between announcement and detail is precisely the territory Brockovich has chosen.

The mismatch is the story. On one side sits a lone activist whose only documented weapon is a demand for transparency. On the other sits a sector announcing commitments measured in tens of billions and gigawatts, moving faster than any disclosure regime tracks. SoftBank's France pledge is one entry in a global race to pour concrete and draw power before the rules catch up.

Brockovich enters that race late, with no regulatory mandate and no named case. What she has is a method that has worked before against companies that assumed no one would ask.

03ClickUp Cut 22% of Its Staff for AI Agents. Box's Founder Says the People Approving Those Cuts Don't Understand the Jobs

ClickUp cut 22% of its workforce this year and routed the work to AI agents. Tech layoffs in 2026 are already approaching the full-year total for 2025, with months still left on the calendar. The executives signing those orders share one conviction: software can now hold jobs that people used to.

Box founder Aaron Levie says those executives are the wrong people to be making the call. On a recent podcast, he argued that whoever decides AI can replace your job is also the least likely to understand what that job actually involves. He filed the behavior under a label now circulating in tech circles: "AI psychosis."

The disconnect Levie describes runs along the org chart. A chief executive watches a demo, sees an agent draft an email or close a support ticket, and scales that up to an entire role. The distance between the demo and the daily work sits with the employees, not the person approving the headcount reduction. By Levie's account, the people most enthusiastic about cutting are the furthest from the tasks being cut.

That framing moved from one founder's opinion to an industry argument on TechCrunch's Equity podcast, which built an episode around whether tech CEOs are "uniquely prone to AI psychosis." The hosts treat it as a debate, not a diagnosis. The question on the table is not whether the technology works, but whether the executive class can judge what it replaces.

Levie offers no clinical claim and no data set, only a structural observation about who holds the authority. ClickUp's 22% is the figure that grounds it. The workers downstream of these decisions get no vote on whether a manager's read of an agent demo matches the job they perform every day.