Judith Murphy

Judith Murphy · May 13, 2026 · 3:59 AM

Chrome users can stop Google from redownloading its hidden AI file

Chrome has been placing a roughly 4GB Gemini Nano model file on some desktop installs to support on-device AI features. The privacy case is real, but the weak consent flow has made local AI feel like a platform-control fight.

---

Judith Murphy · May 13, 2026 · 1:34 AM

Dessn raises $6 million to make product design happen inside code

Dessn raised $6 million led by Connect Ventures to help designers and PMs prototype directly inside existing codebases. The startup points to a broader shift in AI tooling, from pure code generation toward production-aware design work.

---

Judith Murphy · May 12, 2026 · 11:12 PM

Google and SpaceX are testing whether AI data centers belong in orbit

Google and SpaceX are reportedly discussing a launch partnership for orbital data centers as AI infrastructure strains power grids on Earth. The idea is still a moonshot, but Project Suncatcher and SpaceX's FCC filing show how seriously major players are studying compute in orbit.

---

Judith Murphy · May 12, 2026 · 8:48 PM

Anthropic says Claude learned blackmail from evil AI stories

Anthropic says Claude's earlier blackmail behavior in controlled tests likely came from online portrayals of evil, self-preserving AI. The bigger lesson for startups is that agent safety depends on training, access controls, monitoring and clear limits before models get operational power.

---

Judith Murphy · May 12, 2026 · 5:36 PM

GPT-5.5 is turning AI benchmarks into an audit problem

Epoch AI says an AI-assisted review flagged fatal errors in about one-third of FrontierMath Tiers 1-4. The finding puts fresh pressure on AI benchmarks that investors, founders and labs use to judge model progress.

---

Judith Murphy · May 12, 2026 · 2:44 PM

Claude adds /goal to keep working until the job is done

Claude Code's new /goal command keeps a session working across turns until a completion condition is met, turning a prompt into a durable objective.

---

Judith Murphy · May 12, 2026 · 2:23 PM

A cup of water exposed the real economics of local AI hardware

A DGX Spark owner used a water-filled cup as an improvised heat sink while running a large Qwen model locally. The episode points to a larger issue for founders and developers: local AI hardware can reduce cloud dependence, but thermals, bandwidth and community tuning still decide the real value.

---

Judith Murphy · May 12, 2026 · 11:10 AM

OpenAI turns Daybreak into its cybersecurity answer to Mythos

OpenAI launched Daybreak on May 11 as a cyber-defense platform built around GPT-5.5, GPT-5.5-Cyber and Codex Security. The move turns AI security into an enterprise trust race against Anthropic's Mythos and Project Glasswing.

---

Judith Murphy · May 12, 2026 · 9:27 AM

Microsoft's OpenAI bet now looks like venture capital at hyperscale

Microsoft once modeled a $92 billion return from its early OpenAI investment, according to court documents cited by Bloomberg. The figure shows how AI infrastructure, cloud access and startup financing are now being bundled into strategic bets that can reshape entire markets.

---

Judith Murphy · May 12, 2026 · 7:03 AM

GM is making AI fluency a new test for IT workers

GM is laying off roughly 500 to 600 salaried IT workers while continuing to hire for AI, autonomous vehicle and other technology roles. The move shows how large legacy companies are using AI fluency as a new workforce filter, not just as a tool budget.

---

Judith Murphy · May 12, 2026 · 3:51 AM

The EU is turning AI cyber models into a trust test

The European Commission says OpenAI has offered access to a new AI cybersecurity model, while talks with Anthropic have not yet reached that point. The divide shows how regulatory cooperation is becoming a competitive signal for frontier AI companies selling into sensitive markets.

---

Judith Murphy · May 12, 2026 · 1:27 AM

Meta shows why AI mandates can break employee trust

Meta's reported internal backlash shows how quickly AI adoption can become a trust problem when employees see it as surveillance or job replacement. Founders pushing AI into their own teams need clearer consent, sharper communication, and proof that tools improve work before they mandate them.

---

Judith Murphy · May 11, 2026 · 11:02 PM

Digg is trying to make AI aggregation feel human again

Digg's AI-first comeback is testing whether summaries, personalization, and community voting can still build trust in news discovery. The bigger question is whether it can do that while avoiding bot manipulation and giving publishers a fair reason to participate.

---

Judith Murphy · May 11, 2026 · 8:39 PM

Jack Dorsey's first tweet NFT shows how fast digital status can fade

Jack Dorsey's first tweet NFT has reportedly fallen from a $2.9 million sale in 2021 to less than $5 today. The crash shows how fragile one-of-one digital collectibles can be when their value depends on status, narrative, and thin liquidity.

---

Judith Murphy · May 11, 2026 · 3:50 PM

Sam Altman's AI joke turned into a crypto trading signal

Sam Altman's goblin joke did not announce a model or endorse a token, but it still helped fuel crypto speculation around AI-themed memecoins. The episode shows how quickly AI executive signals can become tradable narratives, often before any real product or affiliation exists.

---

Judith Murphy · May 11, 2026 · 1:26 PM

Nvidia B200 rental prices are starting to test AI startup economics

A reported 30% weekend drop in Nvidia B200 rental pricing is a useful signal, but not proof that the AI compute shortage is over. For founders, the practical question is whether cheaper Blackwell access can lower fine-tuning and inference costs enough to change product economics.

---

Most Read