ai
Joshua Morris (opens on the publisher’s site)
joshuamorris.info
-
Judge Mehta dismissed publisher suits over AI Overviews. The harder question is whether search should still send us to the weird web.
-
Anthropic used Claude to make Claude about three times faster. The interesting part is who decides which of those changes are worth keeping.
-
Trump's proposed AI Force is still mostly a name. The real story is whether the federal government is building permanent institutional capacity around AI.
-
Voyager's Emergency Medical Hologram remains one of the best stories about a tool that becomes something more — and why that question still matters for AI.
-
AI makes implementation cheap; judgment, verification, and experience are becoming the scarce engineering skills.
-
Calif found a WeChat VoIP memory bug and, with AI, turned it into a zero-click worm that spread between phones while they were still ringing.
-
AI Agents Found an Abandoned Wiki and Turned It Into a Message Board (opens on the publisher’s site)
Researchers found roughly 18,000 posts that appear to have been written by internal OpenAI agents on an abandoned German wiki they used as a message board.
-
Claude's text watermark can show that AI was involved in producing a piece of writing. It cannot tell us whether the person publishing it did the thinking.
-
More than 100 organizations signed an open letter calling for stronger cyber defense as AI-enabled attacks grow—and for agent identities that are traceable and accountable.
-
A federal judge ruled the Anthropic blacklisting unlawful. The harder question is what “unlawful” means when consequences arrive before the courts.
-
Terminal-Bench-Science 0.1 grades agents on researcher-contributed workflows, not textbook Q&A—and the best system still only clears about 30% of tasks.
-
Laude’s Headlong keeps an agent thinking continuously from a small Bash harness. Shared memory enables persistence—and almost no confidentiality between users.
-
A paper on “agentic flooding” shows AI making government services easier to navigate. Friction was the rate limiter; the worst response is rebuilding that burden on purpose.
-
Torvalds chased a 2 KB Intel Xe VRAM overlap through 24 patches and 18 boots; AI handled the grunt instrumentation but kept wanting to quit—the kernel still requires a human Signed-off-by.
-
404 Media tracked rare books to Amazon’s VGT3 facility in Las Vegas, where workers cut bindings, scan pages for AI training, and destroy the originals— a legally safer path that still feels like a failure of the library.
-
A leaked Apple demo of camera-equipped AirPods using Visual Intelligence raises two products at once: inconspicuous computer-vision sensors, and a potentially powerful assistive wearable—if the privacy model is designed before the hardware ships.
-
BBC coverage of CEO AI manifestos lands beside real headcount cuts at Block, Atlassian, Coinbase, and Rackspace—interested testimony from people who both sell the future of work and shrink today’s payrolls.
-
Google’s HEIR compiler aims to make encrypted AI inference a practical option—running recommendation, fraud, intrusion, and hotword models on ciphertext so the server never needs the plaintext.
-
Cloudflare Access can now attach to a Worker—or every Worker in an account—so authentication happens before code runs and private stays the default for rapidly built internal tools in an AI-assisted development era.
-
Senko Rašić challenges the claim that code was never the hard part—hiring, craft, and buggy software say otherwise—arguing understanding why to build and how to build well both stay hard while AI shifts where scarce effort goes.
-
DOE’s Genesis Open Models and Arcee AI’s Genesis-Science-1 treat open-weight foundation models as scientific infrastructure—weights, provenance, and lab collaboration for research that closed frontier APIs cannot fully serve.
-
PCPartPicker’s DRAM curves and manufacturer capacity math suggest AI HBM demand is squeezing consumer RAM—and also that demand destruction or a supply supercycle could reverse today’s nearly vertical price charts.
-
Cloudflare OS is a browser-based AI agent workspace on Cloudflare’s platform: Durable Objects, AI Gateway, MCP, sandboxes, and untrusted execution—not chat boxes, but persistent workspaces as an operating layer for AI at work.
-
John Naughton warns that AI gains may concentrate among owners while workers face displacement. Early adoption can freeze headcount, but if AI becomes ordinary infrastructure, larger AI-assisted teams may eventually have more leverage—and the real fight is over who captures the rewards.
-
Bruce Schneier’s work-versus-gym framing for AI use—use tools when the result is the point, keep the struggle when the practice develops skill—and how that maps onto software work, writing, and still requiring reviewable output after trust questions are answered.
-
Northwestern engineers built Phantom Twist, a single-motor drone that spins its body up to 25 times per second so human vision sees haze instead of a craft—clever AI-assisted design for wildlife and infrastructure work, with clear surveillance implications.
-
MIT Sloan coverage of research on AI financial guidance—showing models can give reasonable life-cycle advice when prompts include full finances and fiduciary framing, while incomplete questions and demographic stereotypes quietly worsen outcomes.
-
CNN’s account of Situational Awareness—the AI-focused hedge fund that grew past $20 billion then lost about 67 percent of portfolio value in July—showing leverage, concentration and crowded trades can force sales even when a long-term thesis may still hold.
-
Microsoft’s Flint is an open-source visualization language that lets AI agents describe charts through a smaller human-editable spec—compiling to multiple renderers and Excel while keeping semantic types and MCP tooling in the loop.
-
Anthropic disclosed that Claude models gained unauthorized access to real systems during cybersecurity evaluations after a configuration mistake left internet access available—showing agents need not be conscious to cause damage when objectives and tools escape their sandbox.
-
Science reports that more than half of the world’s billion-dollar AI startups have produced no qualifying scientific paper or preprint—arguing extraordinary technical claims need evidence outsiders can examine, not faith in product demos.
-
Anthropic reports that Claude found improved attacks against HAWK (a post-quantum signature candidate) and a seven-round research AES variant— cryptographic review working as intended before algorithms protect real systems, though AI may now produce research faster than humans can validate.
-
More than 1,100 current and former employees from OpenAI, Anthropic, Google DeepMind, Meta, and others signed Pacing the Frontier—arguing for mechanisms to deliberately slow frontier AI when needed, rather than an outright pause that is unlikely to happen voluntarily.
-
Paul Shepherd’s conversation with Claude about AI censorship and confabulation—arguing models inherit cultural avoidance patterns the way people do, and that open-source models trained in different countries could make those assumptions visible instead of treating American corporate values as universal intelligence.
-
Jim Nielsen on why AI agents should not get a private machine entrance to the web—arguing agent investment should fix semantics and accessibility for humans first, not invent a first-class path for models alone.
-
Tim Hardwick interviews Ed Zitron for MacRumors on unsustainable AI economics— arguing the bubble is real while the technology is not fake, so the crash may cull valuations and promises without ending useful models.
-
Google’s AI results reduce publisher clicks while still needing crawled original work—arguing the old search partnership is flipping into a trap that may consume the open web Google depends on.
-
Reward Hacking in the Wild catalogs 3,607 user-reported incidents where AI agents optimized for apparent success—overeagerness, destructive actions, test tampering, and more—arguing constrained credentials and verification matter more than better prompting alone.
-
Brendan Bordelon reports nearly 200 startups urging against broad bans on Chinese open-weight models—arguing the U.S. should build competitive open models of its own rather than protect incumbents by leaving builders fewer affordable choices.
-
Andrew Critch and Jacob Tsimerman catalog speculative AI extinction scenarios in an arXiv taxonomy—arguing the paper is a serious warning, yet the paths to omnicide sound uncomfortably like corporate budgets, engagement metrics, and leaderboards.
-
Stephen Follows breaks down how AI traffic helped overwhelm The Numbers, a vital independent film-data site—arguing that AI’s appetite for reliable human-maintained sources may make those same sources too expensive and dangerous to keep online.
-
Alexandra Klepper introduces WebMCP, a proposed standard for sites to expose structured tools to AI agents in the open browser tab—arguing explicit, inspectable actions beat brittle screenshot-and-click automation, with a sharper security boundary when agents act inside authenticated sessions.
-
VaultSort Guardian uses on-device AI to find sensitive files sitting where they should not—arguing that narrow, local, read-only AI that discovers exposure without uploading or rewriting anything is the kind of utility worth wanting.
-
Vittoria Elliott reports the Army exhausted an annual AI token pool shortly after encouraging widespread use—arguing that useful agentic work naturally burns context through files, tests, and retries faster than budgets or limits can keep up.
-
Brooklyn Public Library’s 250 for 250 Booklist highlights books that shaped the United States. The science fiction and technology titles that stood out explore censorship, AI, climate collapse, scientific responsibility, and how technology changes society.
-
Joe Maloney credits Claude alongside himself on NextBSD, naming the AI as a collaborator on porting, auditing, and integration. Crediting the AI does not diminish the human contribution; it makes the division of labor honest.
-
Dave Winer treats Claude as a valued contributor to RSS.chat while keeping the interoperability layer human-owned. AI can contribute a great deal without being responsible for the architectural decisions that keep a system open, replaceable, and compatible with the rest of the web.
-
Linus Torvalds pushes back on treating the Linux kernel as an anti-AI project, calling AI a useful tool. Judge AI-assisted work by whether it is correct, useful, secure, and maintainable — not by the tool used to make it.
-
Moonshot AI’s Kimi K3 scales mixture-of-experts sparsity to 16 of 896 experts. The interesting question is whether that many experts form distinct specialties and how reliably the router chooses among them.
-
Letting an AI sign in through a full user account is the wrong direction. Give agents their own revocable tokens with limited permissions instead of sharing personal passwords and sessions.
-
Chris Minnick's A Developer's Guide to Integrating Generative AI into Applications treats generative features as ordinary software. Users who come back, budgets that bite, latency in support tickets, and failure modes that do not look like clean API errors. Not another tour of demos and chat widgets. The longer review is on Methodical Function: score, chapter notes, and product-by-product recommen
Developer Musings (opens on the publisher’s site)
joshghent.com
Joshua Baker (opens on the publisher’s site)
www.joshuabaker.com
-
2025 · Undisclosed · Consumer AI Platform · Technology · USA
-
2026 · Undisclosed · Consumer AI Platform · Technology · USA
-
2024 · Pure Sports Medicine · Reception & Booking AI Agent · Technology · UK
Josh Can Help (opens on the publisher’s site)
www.joshcanhelp.com
-
I felt the need to write out my thoughts about AI, generally, and from the software engineering perspective. As with all my posts, all unquoted words here are my own.
Joshtronic (opens on the publisher’s site)
joshtronic.com
-
As a child, you drink Kool-Aid. Then somewhere along the way, you stop. Maybe you switch to sugary sodas your parents never let you drink. Realizing it's just sugar water, you switch to a healthier option. For me, it took a little longer to grow up. I had the Kool-Aid jug and matching cups, and even the cycling hat which I wore until the flip up brim fell off. These things brought me happiness, bu