AI is Now Visible in Productivity Statistics

February 17, 2026 · Episode Links & Takeaways

Author

AI Daily Brief
February 17, 2026

HEADLINES

Anthropic vs. The Department of War

Anthropic's standoff with the Pentagon is boiling over. The Wall Street Journal reported Claude was used in the Maduro raid, and Anthropic appears to have reached out to Palantir — which serves Claude to the military — to find out exactly how. A senior administration official told Axios the inquiry was "raised in such a way to imply that they might disapprove of their software being used, because obviously there was kinetic fire during that raid, people were shot." Anthropic denied discussing Claude's use in any specific operations.

The Department of War responded with a pointed statement from spokesperson Sean Parnell: "Our nation requires that our partners be willing to help our warfighters win in any fight." Fox News reported Monday that Secretary Hegseth is "close" to cutting ties entirely and banning contractors from using Anthropic models. A senior DoW official called Anthropic a potential "supply chain risk" — a designation Axios noted is usually reserved for foreign adversaries, with the Huawei ban as the clearest comparison. Another official told Axios: "We are going to make sure they pay a price for forcing our hand like this." This isn't just an Anthropic story — it's a proxy fight over who gets to dictate the terms of AI's use: the companies that built it, the governments they operate in, or some constantly negotiated combination.

Alibaba Launches Qwen 3.5 Ahead of DeepSeek V4

Another Chinese model slips in ahead of the anticipated DeepSeek V4 release. Alibaba's Qwen 3.5 Plus has 397 billion parameters in a mixture-of-experts architecture with a million-token context window. Benchmarks put it broadly in line with GPT-5.2, Opus 4.5, and Gemini 3 Pro — a big improvement over Qwen 3 Max Thinking but short of the latest US frontier. The big new capability is native multimodal reasoning, making Qwen and Kimi K2.5 the only open source models with this feature. Alibaba Cloud is serving it at $1.20/$7.20 per million tokens — even cheaper than Kimi K2.5. It continues to reinforce just how much intelligence is available for ever-diminishing prices from the Chinese labs.

Hollywood in Full Freakout Mode Over Seedance 2.0

When ByteDance released Seedance 2.0 last week, director Ruairi Robinson showed what it could do with a 15-second clip of Tom Cruise fighting Brad Pitt in a post-apocalyptic cityscape. Deadpool writer Rhett Reese told the New York Times the video "sent a cold shiver up his spine" — "I could just see it costing jobs all over the place." The core realization: Chinese AI firms are not going to play ball on copyright. The MPA denounced "unauthorized use of U.S. copyrighted works on a massive scale," and SAG-AFTRA called it "unacceptable." ByteDance acknowledged the criticism, saying they'll "strengthen current safeguards" — but even if they do, whether smaller Chinese labs play by those rules seems extremely unlikely. Still, there's a growing sense that Hollywood will just have to adapt. Reese himself acknowledged: "Hollywood has long been a gatekeeper that keeps young and poor people away from creative levers. When a young person with no capital sets out to impress Hollywood, they will use tools like these. And young Chris Nolans will be among them."

Apple Announces Mysterious Multi-City Event on March 4

Apple has invited press to New York, Shanghai, and London for a "special Apple Experience" on March 4 — unusual given they typically hold major releases around WWDC in June. Bloomberg's Mark Gurman pointed to a big hardware lineup: new MacBook Pros and Airs with the M5 chip (30% unified memory bandwidth boost over M4), a low-cost colorful MacBook, new iPads, and the iPhone 17e. For AI folks, the big question is whether this includes a surprise M5 Mac Mini — if so, expect a lot of OpenClaws shopping for an upgrade. The other question is AI Siri: Tim Cook has said the new Siri ships this year, but most expect it at WWDC. Behind the scenes, Gurman reported the project is hitting roadblocks with query processing and response speed, though March was the original target and things remain fluid.

MAIN STORY

AI is Now Visible in Productivity Statistics

The single big overarching theme of 2026 so far is the sense among AI builders and power users that a key inflection point has been hit. Alongside that has come an invigorated conversation about AI's impact on workers — but it's still largely rooted in anecdote. As the discussion gets more urgent, it becomes more important to move out of the paradigm of anecdotes and into actual evidence. Stanford economist Erik Brynjolfsson now believes he's seeing the first real sign of AI productivity showing up in the macro data — and if he's right, it would be a major break from how every prior technology paradigm has played out.

PRODUCTIVE DEBATES ABOUT AI PRODUCTIVITY

Andrew Yang and the Anecdote Problem
"This will result in the great disemboweling of white-collar jobs."
The AI job displacement conversation has ratcheted up significantly this year, but it's still largely rooted in anecdote. Andrew Yang's latest blog post "The End of the Office" exemplifies this: he watched a family member build a website with AI in minutes and extrapolated to the elimination of all white-collar work. The urgency is real, but as the conversation intensifies, the gap between anecdote and evidence becomes more important to close.

Andrew Yang Blog The End of the Office

The Solow Paradox and Why This Matters
"You can see the computer age everywhere, but in the productivity statistics."
That famous Robert Solow quote from 1987 came after decades of IT investment. We eventually saw a productivity jump in the late '90s, but nothing comparable in the Web 2.0/mobile/social era — productivity growth has actually been below the historical pattern for the last 20 years. If AI is already showing up in productivity numbers, that would be a major departure from how every prior general-purpose technology has played out — and a major indicator of just how different this shift really is.

Brynjolfsson's Argument — The J-Curve Inflection
"We are now transitioning out of this investment phase into a harvest phase."
Brynjolfsson points to the BLS revising 2025 job numbers down by ~400,000 (from 584K to 181K net gains) while GDP remains incredibly strong — Q3 at 4.4%, Q4 provisionally at 3.7%, with the Atlanta Fed's GDP Now even higher at 5.4%. Since productivity is GDP divided by workers, the downward jobs revision implies productivity growth of roughly 2.7% for 2025 — almost double the prior decade's average. Brynjolfsson argues this solves the conundrum posed by Apollo Chief Economist Torsten Slok, who wrote: "AI is everywhere except in the incoming macroeconomic data." Brynjolfsson's J-curve thesis — that general-purpose technologies require massive upfront intangible investment before benefits are "harvested" — dates back to a 2018 paper. He now believes we're reaching the inflection point.

Erik Brynjolfsson (X) Productivity growth

The Pushback — "Very Thin Evidence"
"May turn out to be true but it's very thin evidence."
Not everyone agrees. Economist Guy Berger warned against drawing strong inferences from this data point. He pointed to the actual revised statistics: by and large, the revision was about removing government workers from DOGE cuts plus counting layoffs in mining, logging, transportation, and manufacturing — not the AI-exposed white-collar sectors Brynjolfsson's thesis would predict. Noah Smith and Chicago economist Alex Imas were more enthusiastic ("I guess 'sooner' came pretty quickly"), but the correlation between these specific job losses and AI exposure is questionable.

The White-Collar Recession Is Real Regardless
"The US white-collar recession is accelerating."
Whatever's actually driving the numbers, the white-collar labor market is brutal. The Kobeissi Letter reported just 1.6 job openings per 100 employees in professional and business services — the lowest in 11 years, lower than the 2020 pandemic bottom, with hiring rates matching 2008 financial crisis levels. Politicians on both sides are taking notice. Republican Jay Obernolte (who has a master's in AI) said: "There will be job displacement. We need to re-skill the workers." Democrat Elizabeth Warren warned: "If AI comes in on top of that and literally wipes out the income for millions of families, we're going to see a full-blown crisis."

The Canaries Paper Follow-Up
"Some of the earlier declines are likely due to a combination of factors, not just AI."
Brynjolfsson and colleagues circled back to their "Canaries in the Coal Mine" paper with a new note addressing two key criticisms: whether rising interest rates better explain the hiring decline, and whether the timing matches AI exposure. They found interest rates don't explain the disproportionate drop in entry-level hiring in AI-exposed occupations, but also acknowledged that the employment effects only become statistically significant in 2024 — earlier declines were likely multi-causal.