Tech Digest – March 9, 2026
Workforce Disruption
Block Cuts 4,000 Jobs, Blames AI — Analysts Call It “AI-Washing”
Block, Jack Dorsey’s payments company behind Square and Cash App, laid off more than 4,000 employees — nearly half its workforce — citing AI-driven efficiency gains. Dorsey told shareholders a smaller team “using the tools we’re building, can do more and do it better” and predicted most companies would make similar cuts within a year. Block’s stock surged 17–24% on the announcement. But Bloomberg, Wharton’s Ethan Mollick, and multiple analysts flagged the move as potential “AI-washing” — dressing up a correction for pandemic-era overhiring (Block tripled headcount from 3,800 in 2019 to 10,000) as technological inevitability.
Note: The market rewarded the cut before anyone could measure whether AI actually replaced the work. That sequence — announce AI efficiency, cut staff, stock jumps — creates a template other CEOs will be tempted to copy. For institutions watching this as a workforce planning signal: the question isn’t whether Block’s AI claims are real, it’s whether other employers will act as if they are.
New Study Identifies “AI Brain Fry” — The Cognitive Cost of Supervising Machines
A study of 1,488 U.S. workers published in Harvard Business Review identified a condition researchers call “AI brain fry” — a mental fog caused by prolonged AI oversight. Symptoms include buzzing sensations, slower decision-making, and headaches. The finding highlights a counterintuitive problem: the more AI handles execution, the more cognitive load shifts to the humans monitoring it.
Note: “Automate the task, supervise the output” sounds efficient until you measure what supervision actually costs. Any institution deploying AI copilots should be budgeting for the human overhead, not just the software license.
Sources: Harvard Business Review
Adoption at Scale
Over Half of UK Adults Now Use AI for Financial Advice
A Lloyds Banking Group study of 5,000 Britons found that more than half now use AI tools for financial advice, with one in three consulting AI weekly. The shift is happening ahead of any regulatory framework governing AI-generated financial guidance — citizens are adopting the tools faster than institutions can set guardrails.
Note: Citizens aren’t waiting for permission. When more than half the public uses AI for consequential decisions, the question for regulators and service providers isn’t whether to engage — it’s how far behind they already are.
Sources: Financial Times
Shenzhen Treats AI Agent Adoption as Municipal Infrastructure
Shenzhen’s Longgang District released ten draft policy measures supporting OpenClaw, an open-source AI agent platform that has gone viral in China. The package includes subsidies up to 2 million yuan (~€290,000) for approved applications, free compute credits, discounted office space, and financing up to 10 million yuan for seed-stage projects. The district is explicitly targeting “one-person AI companies,” framing agent infrastructure as economic development. Chinese cloud stocks rose at least 9% on the news, and Tencent hosted a free installation event in Shenzhen that drew crowds ranging from children to retirees.
Note: A local government subsidizing agent deployment the way others subsidize broadband or industrial parks. That’s not an AI experiment — it’s infrastructure policy. European municipalities competing for tech investment should note what “AI-ready” looks like in Shenzhen: not a strategy document, but compute credits and office space.
Sources: South China Morning Post, Dataconomy
Research Automation
Karpathy’s “Autoresearch” Runs 650 Experiments in Two Days — Improvements Transfer Across Model Scales
Andrej Karpathy reported that his “autoresearch” project — an autonomous AI system tasked with discovering training improvements — ran approximately 650 experiments over two days and found gains that transfer from smaller to larger model architectures. The system autonomously designed, executed, and evaluated experiments without human intervention, prompting Karpathy to note: “Who knew early singularity could be this fun?”
Note: AI improving AI training is no longer theoretical — it’s running on someone’s laptop and posting results to X. The feedback loop between AI capability and AI research speed is now measurable. Timelines for “what AI can do next year” just became harder to predict from the outside.
Sources: Andrej Karpathy (X)
Physical AI & Robotics
London Surgeon Performs UK’s First Remote Operation — 2,400 km Away in Gibraltar
A London-based surgeon performed the UK’s first long-distance robotic operation on a patient approximately 2,400 km away in Gibraltar, describing the experience as feeling “almost as if I was there.” The procedure demonstrates that telesurgery is moving from proof-of-concept toward clinical deployment, with implications for healthcare access in underserved regions.
Sources: BBC
UK Company Deploys Portable Micro-Factories for Robotic House Construction
AUAR, a UK-based construction technology company, is deploying portable micro-factories that produce wooden house framing faster and at lower cost than human crews. The robotic systems handle precision cutting and assembly of structural components, freeing skilled carpenters to focus on final assembly and finishing work rather than repetitive framing tasks.
Note: Housing crises across Europe share a common bottleneck: not enough skilled labour to build fast enough. A portable factory that ships to the construction site — rather than requiring workers to come to it — is exactly the kind of infrastructure solution municipalities should be tracking.
Sources: CNN
Financial Infrastructure
Stablecoin Payments for AI Agents Take Shape as Nasdaq Moves to 24/7 Trading
Circle, Stripe, Coinbase, and other fintech firms are building stablecoin-based payment infrastructure designed to make microtransactions between AI agents economical — enabling autonomous software to pay for services without human intervention. Separately, Nasdaq announced a partnership with Kraken to offer tokenized 24/7 stock trading by early 2027. Box CEO Aaron Levie framed the shift bluntly: developers should stop building software for humans and start building for “trillions of agents” taking over enterprise work.
Note: Financial infrastructure is reorganizing around two assumptions: markets never close, and most transactions won’t involve a human. Procurement frameworks, budget controls, and audit trails designed for human-speed, business-hours operations will need rethinking.
Sources: Bloomberg, Wall Street Journal
Legal & Biotech
Federal Judges Rule AI Bots Are Not Persons and Lack Legal Standing
In two recent U.S. federal cases, judges effectively declared that AI systems are not human, do not possess rights reserved for people, and produce outputs that do not merit special legal standing. The rulings begin to establish precedent around AI personhood — or rather, the absence of it — at a time when autonomous agents are entering commercial and public-sector workflows.
Note: As AI agents start signing up for services, booking meetings, and transacting autonomously, every institution will need a position on what an agent can and cannot do in their name. These rulings suggest the legal system is drawing that line early: agents are tools, not actors.
Sources: Yahoo News
GLP-1 Drugs Linked to 14% Reduction in New Substance Use Disorders
A large-scale study of U.S. veterans found that GLP-1 receptor agonists — the class of drugs behind Ozempic and similar treatments — reduced the likelihood of developing new substance use disorders by 14% and sharply cut overdose rates among those already affected. The findings add to a growing body of evidence that GLP-1 drugs have effects well beyond weight management.
Note: A single drug class that affects obesity, cardiovascular risk, and now addiction simultaneously is a health policy wildcard. For any institution managing workforce health programs or public health budgets, the cost-benefit calculus on GLP-1 coverage just changed again.
Sources: Science
