Briefings

Morning Briefing — Friday, February 27, 2026

Block's AI-native restructuring reshapes fintech

Block fires 4,000+ humans to become "AI-native." The coding agents Karpathy says "basically work" are already replacing their colleagues.

💼 Economics & Labor

Block Cuts 40%+ Workforce to Become AI-Native Fintech SIG 4
Hacker News
Jack Dorsey's Block announced sweeping layoffs cutting headcount from over 10,000 to under 6,000 — a 40%+ reduction. The company is restructuring to become an "AI-native" fintech, making it one of the largest AI-driven workforce reductions in tech history. This isn't a quiet efficiency play — it's a public declaration that AI can replace nearly half your engineers. Every fintech CEO is watching.
SF Fed: AI Activity Adds Uncertainty to 2026 Economic Outlook SIG 4
San Francisco Federal Reserve
The San Francisco Fed's February FedViews highlights AI-related activity as a key source of uncertainty in the 2026 economic outlook, particularly for knowledge-intensive industries. Significant disruptions to employment patterns in tech, finance, and professional services are cited. When the Fed starts naming AI as a macroeconomic variable alongside inflation and rates, the disruption has left the speculation phase.
Will Vibe Coding End Like the Maker Movement? SIG 3
Technically / Hacker News
An essay comparing the "vibe coding" trend to the maker movement hit #4 on HN with 320+ comments, questioning whether AI-assisted coding will fizzle out similarly. The deep developer community debate reflects real anxiety: is this a democratization moment or a hype cycle? The maker movement comparison is apt in some ways (lowered barriers, early enthusiasm) but misses a key difference — 3D printers didn't keep getting better every quarter.
JAMA Study: AI Use Linked to Increased Depressive Symptoms SIG 3
JAMA Network Open
A peer-reviewed study in JAMA Network Open finds an association between generative AI use and depressive symptoms among US adults. Trending on Hacker News, this is the first major clinical study to quantify psychological impacts of AI interaction. Correlation isn't causation, but it's enough to fuel regulatory conversations about AI mental health effects — especially as AI companions proliferate.

🤖 Agents & Tools

Karpathy: Coding Agents "Basically Work" Since December SIG 4
Andrej Karpathy
Andrej Karpathy posted that programming has fundamentally changed in the last two months. Coding agents "basically didn't work before December and basically work since" — models now have high enough accuracy that iterative agent loops reliably converge on solutions. Coming from Karpathy, this isn't hype; it's a timestamp on when the shift happened. Pair this with Block's layoffs above and the SF Fed warning — the labor market implications of "coding agents basically work" are being priced in real-time.
Alibaba Releases OpenSandbox: General-Purpose AI Agent Platform SIG 3
GitHub Trending
Alibaba open-sources a general-purpose sandbox platform for AI applications with multi-language SDKs, unified APIs, and Docker/Kubernetes runtimes. Supports coding agents, GUI agents, agent evaluation, and multi-agent workflows. Another sign that the agent infrastructure layer is commoditizing fast — if Alibaba is giving away the plumbing, the value moves up the stack to orchestration and domain expertise.

🔬 AI Research & Security

WiFi DensePose: Full-Body Pose Estimation Through Walls SIG 4
GitHub Trending
A production-ready implementation of InvisPose — a WiFi-based dense human pose estimation system that enables real-time full-body tracking through walls using commodity mesh routers. Trending #1 on GitHub. The surveillance implications are staggering: any WiFi router could theoretically become a through-wall body tracker. This was academic research 18 months ago; now it's a GitHub repo with install instructions. The gap between "technically possible" and "anyone can do it" just closed.

🛡️ Robotics & Defense

AI Flies on F-35 in Project Overwatch Breakthrough Test SIG 3
The Defense Post
Lockheed Martin integrated AI-enhanced Combat Identification into the F-35's information fusion system during Project Overwatch testing at Nellis AFB. The system resolves overlapping signals and ambiguous radar returns to identify threats faster than human operators. While the Pentagon fights Anthropic over Claude access, the defense-industrial complex is quietly embedding AI into actual weapons platforms through other channels.
Alphabet's Intrinsic Robotics Joins Google for DeepMind & Gemini Integration SIG 3
American Bazaar
Alphabet-owned robotics platform Intrinsic is being absorbed into Google proper, gaining direct access to DeepMind research and Gemini models for industrial robotics. This consolidation signals Google is getting serious about physical AI — combining its best research lab, its best foundation model, and its robotics division under one roof. The race to bring LLM-grade intelligence to physical systems is accelerating.

🔭 Secretary's Assessment

Today's briefing tells one story from three angles: the AI labor shock is no longer theoretical.

Block just made the quiet part loud. Cutting 40% of your workforce and explicitly calling it an "AI-native" restructuring is a message to every board in Silicon Valley. This isn't "efficiency gains" or "strategic realignment" — it's a CEO saying the machines can do what half his employees did. Pair this with Karpathy's timestamp ("coding agents basically work since December") and the SF Fed flagging AI as a macroeconomic uncertainty, and you have three independent signals converging on the same conclusion: Q1 2026 is when AI labor displacement moved from speculation to execution.

The WiFi DensePose story deserves more attention than it will get. Full-body pose estimation through walls using commodity hardware was a research curiosity in 2024. Today it's a trending GitHub repo. This is the pattern we keep seeing: capabilities that seemed safely theoretical become democratized overnight. The privacy implications are enormous — every mesh WiFi system in every home is a potential surveillance device. Expect this to surface in policy discussions within weeks.

The defense stories are a quiet counterpoint to the Anthropic saga. While the Pentagon threatens Anthropic over Claude access, Lockheed Martin is already flying AI on F-35s through Project Overwatch, and Google is consolidating its robotics under DeepMind. The military doesn't actually need Anthropic — it has plenty of willing AI partners. The ultimatum was always more about establishing precedent (can the government compel AI companies to provide unrestricted access?) than about operational necessity. Today is the deadline day. We'll know by tonight whether Anthropic blinked.

Watch for: Anthropic's response to the Pentagon deadline (5pm ET today). Block's stock movement as markets digest the AI-native pivot. And whether other companies follow Block's lead in framing layoffs as AI transformation rather than cost-cutting.