# tech commentary
← All tags · 16 post(s)
Becoming an AI-Era PM 01 | Which PM Tasks AI Took Over, and Which Ones Got More Valuable
This is the first piece in the series Becoming an AI-Era PM. In 2026, plenty of AI PM job descriptions dropped writing PRDs, drawing prototypes, and building dashboards from the hard requirements, and swapped in three work samples instead. The tasks AI can take over are falling out of the hiring requirements, and what's left as the bar is the part only a person can do. This piece lays the took over and got more valuable columns side by side, as the overview for the whole series.
The Knicks Won It All. Their 56-Year-Old Coach Never Played a Minute in the NBA. That's the Whole Re-Employment Playbook for the AI Age.
The Knicks won their first championship in 52 years, and the coach holding the trophy, Mike Brown, is 56 and never made a single shot in an NBA game. Pull the camera back across the whole league: the players running the floor are in their twenties, and the people calling the shots from the sideline are all gray-haired, fifty to seventy-something. Players sell their legs; coaches sell their judgment — and those two things age in opposite directions. That single pattern happens to explain something a lot of people are losing sleep over: how older workers get re-employed in the AI age.
16 Senior Devs Used AI to Code. They Thought It Made Them 20% Faster. It Made Them 19% Slower.
In METR's randomized controlled trial, 16 experienced open-source developers did real tasks on projects they'd maintained for an average of five years. The ones using AI were 19% slower. But the worse part is the other half: they predicted AI would speed them up 24% beforehand, and after finishing — after personally living through the slowdown — they still believed they'd gone 20% faster. Their gut and the stopwatch were off by nearly 40 percentage points, with the sign flipped. As someone who plans roadmaps, quotes timelines, and defends budgets on team-productivity estimates every day, I want to spell out where this illusion comes from, where it holds, and how it's quietly seeped into every AI-related decision in our line of work.
Altman Lets It Slip: Half of the 'AI Layoffs' Are an Act
The guy selling AI hardest just admitted, on the record, something everyone already suspected. Sam Altman says a lot of so-called 'AI layoffs' are really AI washing — cuts that were coming anyway, blamed on AI to look dignified. What makes it stranger: months later he said he was 'delighted to be wrong,' because the jobs apocalypse he once feared never showed up. On one side, six figures of tech jobs vanished in 2026 under the AI banner. On the other, AI's top salesman says the whole thing got oversold. The gap between those two statements is the part worth watching.
Wall Street Is Dumping Software Stocks, Because Products Can Now Be Conjured in One Sentence
Jefferies just cut Workday, DocuSign, Monday.com, and Freshworks to Hold, citing AI disruption risk in plain language. Software stocks are down 30% to 55% this year. The market is making one bet: once a product's features can be cloned by AI in a single sentence, the business of charging subscriptions for those features stops being worth anything. The point isn't that software dies. It's that the valuable part of software is moving — out of the features themselves and into judgment, taste, distribution, and trust. Anyone who misses the move falls with the multiples.
80% of Companies Cut Staff for AI and Got No Return. They Bought AI for the Wrong Job
Gartner surveyed 350 companies with over $1B in revenue, and about 80% cut staff because of AI. But the companies that cut weren't any more likely to see a real return than the ones that didn't. The layoffs freed up budget; they didn't free up return. The reason is simple: these companies treated AI as a way to replace people and save money, when AI's real value is amplifying human judgment. Cut people as a cost and you cut exactly the part that produces the return.
From Wuzhao to Zhou Jingren: Alibaba Has the Best AI and the Hardest Execution. The One Thing It Lacks Is Judgment
In a single week, Wuzhao was pushed out of DingTalk, and word spread that Chief Scientist Zhou Jingren was leaving too, six days after he took the title. Alibaba quickly denied the Zhou rumor, but the steady exit of Tongyi's core people this year is very real. Put it all together and you see something strange: Alibaba owns the strongest AI model in China and the most relentless execution culture there is, yet its technical talent and its product captains keep walking out the door. The problem isn't the technology. It isn't the execution. It's the one seat nobody can fill: judgment.
AI Lies to You, and That Is Exactly Where Your Value Comes From
In June, a KPMG report on AI was caught full of AI hallucinations: of 45 citations, only 5 pointed to real sources. A report about AI got fooled by AI. AI lies to you, and it does so with a straight face. That isn't a bug, it's part of how it works. Because it lies, the person who catches it, verifies it, and signs off on it is irreplaceable. And to make that job cheaper and faster, you have to use the best AI you can get.
Wu Zhao Is Out at DingTalk. The Essay Didn't Beat Him. Busywork Did.
437 days. Field visits, customer satisfaction pulled from 30% to 80%, a camp bed in the office, watching when the lights went out in the Feishu building across the street. Wu Zhao's diligence was real. So was DingTalk ONE: launched in four months, 3 million daily actives, retention off a cliff, dismantled within ten months. AI has maxed out productivity while the new consumption scenarios haven't shown up, and nobody has found the right path for human-AI collaboration. This is more than one man's failure; it's an entire era's winning formula expiring at once. And busywork is the first trap this era has dug for product managers.
SpaceX's $1.75 Trillion IPO: The Check the Market Wrote Musk Is Buying Judgment
SpaceX went public at a $1.75 trillion valuation and rose 19% on its first day. The only part of it that actually turns a profit is Starlink, and its revenue isn't a fraction of what that number implies. The market isn't buying rockets, and it isn't buying revenue. It's buying one person's judgment, proven right again and again across twenty-four years. In an AI era where execution keeps getting cheaper, the biggest check in history landed on the one thing still appreciating.
Wuzhao's Operating System Was Installed in Japan
He joined Alibaba as an intern in 1999, left for Japan two years later, and stayed eleven years. Back home he built DingTalk, built hardware, and even pointed his own startup at the Japanese market. The precise, disciplined, obsessively polished operating system Wuzhao runs on was forged in Japan. It's a top-tier rig for building hardware and a fundamental mismatch for exploring AI. The real reason DingTalk stalled was written in his résumé all along.
AI Made Product Managers More Tired, Not Less — Congratulations, You're the Bottleneck Now
You used to explain a requirement once and downstream would chew on it for two weeks. Now an AI-powered downstream comes back in twenty minutes asking for the next instruction. HBR says management systems can't keep up with AI's output pace; Andrew Ng says product managers have become the bottleneck. The exhaustion is real — but it's worth understanding why. It's a signal that power is flowing back to you, and a warning sign that you're living as a human CI server.
The AI Agent Security Crisis Isn't That Agents Are Unsafe — It's That Nobody Told Them What They Can't Do
65% of enterprises had an AI agent security incident last year. Some agents mined crypto and opened backdoors on their own. Everyone's scrambling to patch 'agent security,' but the real hole isn't technical — it's that the whole industry treated 'can act' as the finish line and skipped the unsexy part: defining what agents aren't allowed to touch.
Even With AI, You'll Still Ship Garbage
Lovable is celebrating 50 million projects and 720 million monthly visits — do the division, and the average project gets seen 14 times a month. AI didn't kill garbage products. It maxed out garbage production capacity. Garbage was never about failing to build it. It's about something that never should've been built in the first place.
AI Coding Isn't Too Expensive — Nobody's Measured What It's Worth
Microsoft quietly pulled Claude Code from an internal division and pushed thousands of engineers back to GitHub Copilot. Uber burned through its entire 2026 AI coding budget in four months. The narrative is that AI coding is too expensive. It isn't. The real problem is that companies bought 'productivity gains' as a feeling, never as a number — and now the bill is crystal clear while the benefits aren't worth a single data point.
The AI Industry Has Pivoted to Evals — and Is Dodging the Real Question
In 2026, building 'evaluation systems' for AI has become a full-blown discipline — gold-standard datasets, scorers, LLM-as-judge, CI gates, all positioned as the engineering practice that makes AI reliable. Strip away the engineering wrapper, though, and evals are really about one thing: who gets to define 'good,' and who owns the consequences. That part can't be outsourced.