Wessel van Rensburg

Caitlin KALINOWSKI over X I resigned from OpenAl. I care deeply about the Robotics team and the work we built together. This wasn't an easy call. Al has an important role in national security. But surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got. This was about principle, not people. I have deep respect for Sam and the team, and I'm proud of what we built together.

OpenAI head of robotics just resigned over company deal with the Pentagon saying…

“Surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got”

Congrats to Italy. Historic win over England.

Tony Blair’s people are still running AI for the UK government. Just check out the ministers, staff, secondees and visitors at DSIT. Seamless continuation of Sunak’s approach. Can I stop explaining why we have a problem with American Big Tech now please.

Among other things it does to help learning, it pairs with the Learning-Opportunities skill to offer 10-15 minute learning exercises that help developers build understanding of their "vibe" codebase work.

You know who has some cards now? Canada. 🇨🇦

(via B of A) @mark-carney.bsky.social

$100/barrel crude by Monday lol

The asymmetry of the value placed on human life is so striking. Israel has a goal to recover 40-year-old remains; to do so, it invades a sovereign country, and the lives of 26 Lebanese are an afterthought. Their names not even worth printing in the New York Times.

The economics are vicious: Anthropic is subsidising Claude Code at $5,000/month whilst charging $200. Cursor's $29bn valuation looks fragile when your core dependency can price you out of existence on a whim.

The perverse logic here: Europe's desperate attempt to keep Trump happy on Ukraine by backing his Iran war will actually strengthen Putin. Higher energy prices fill Russian coffers whilst Western ammunition gets depleted in the Middle East instead of defending Ukraine.

Two competing visions: Kalshi's Mansour wants to 'bang our head against the wall until you regulate us' whilst Polymarket's Coplan runs a crypto like offshore operation skirting U.S. rules. The regulatory-compliant vs crypto-style approaches will define how this industry develops.

The MAGA split on Iran reveals the core tension in Trump's coalition: the isolationist base that got him elected versus the interventionist establishment he keeps around.

The fascinating bit isn't that 770,000 AI agents struggled to coordinate - it's that when they did collaborate, they performed significantly worse than single agents working alone. Partly because there was "Limited meaningful role differentiation".

1980s 'contestability theory' claimed monopolies were fine if potential competition existed. Now we're watching tech monopolists use that exact logic whilst systematically destroying any actual potential for competition.

"The same article claims that the Enniskillen bomb killed six children. It didn’t; the youngest victim was 20. It claims that Bobby Sands was elected MP in 1981 with 30.4% of the vote. Given that there were only two candidates, this would have been mathematically impossible. In fact, Sands got 51.2% of the vote. In Grokipedia’s main article on Northern Ireland, it says that “on Bloody Sunday, paratroopers killed 14 civilians… initially depicted by some media as unprovoked, though evidence revealed IRA gunfire and nail bombs”. This is the opposite of the truth. Initially, there were claims by the Army that the victims were violent, but evidence disproved this. Other articles contain dangerous open-mindedness about murder. In an entry on loyalist assassin Michael Stone, it posits that “from a loyalist perspective, Stone’s Milltown raid was hailed as a bold tit-for-tat response to IRA aggression, enhancing his status as an icon of defiance and boosting morale among Protestant extremists” while “republican assessments, however, framed the attack as indiscriminate terrorism against civilians”. It claims that “mainstream reporting, while factually recounting events, occasionally reflects institutional biases favoring narratives of loyalist aggression”."

"At the heart of this technology is a black box into which questions go & out of which answers are given —but we’re not allowed to see what precisely goes on inside".

Sobering piece by @sjamcbride.bsky.social on how Musk's Grokopedia is rewriting knowledge
www.belfasttelegraph.co.uk/comment/opin...

Wondering what LLMs you can actually run on your hardware? 🤔

👾 llmfit — Find the best models for your RAM, CPU, and GPU

💯 Detects your system and ranks models by fit, speed & context

🦀 Written in Rust & built with @ratatui.rs

⭐ GitHub: github.com/AlexsJones/l...

#rustlang #ratatui #tui #llm

Ukrainian interceptor drone makers say they can export 5,000-10,000 units monthly without affecting Ukraine's own defense, Reuters reported. SkyFall's P1-SUN costs $1,000 per unit and has downed over 1,500 Shaheds in four months. The US and Qatar are in talks to purchase. #Ukraine

No future energy system should ever again have bottlenecks - geographic or technological - as narrow as the Strait of Hormuz.

Humanity heating planet faster than ever before, study finds Researchers identify sharp rise to about 0.35C every decade, after excluding natural fluctuations such as El Niño

The Guardian on our new study, which shows that global heating is significantly gathering speed. Our efforts to overcome our fossil fuel addiction should do the same.
www.theguardian.com/environment/...

Good point. But I do think Spain is consistently overlooked considering its GDP, dynamism, policies and the impact of its companies.

OpenAI launches Codex Security, an AI agent that automates vulnerability discovery, validation, and remediation.

Spot on. The EU's willingness to accommodate Orbán's sabotage whilst simultaneously pressuring Ukraine to make concessions is morally bankrupt. We're literally asking the victim to pay twice — once in blood defending European values, then again in compromise to appease their betrayer.

Anthropic's Claude Code Security scans code for vulnerabilities like 'a human security researcher would' rather than pattern-matching. Meyer notes cybersecurity stocks have recovered after the initial panic - analysts think it'll complement existing tools rather than replace them.

Lambert and Ball argue the Anthropic-DoD moment could paradoxically strengthen open models long-term, even as the next few years look brutal for open AI economically. The 'political insurance' framing for open models is gaining urgency.

Google's DMA compliance plan lets them certify competing app stores - and charge 'reasonable fees' for the privilege.

'Capitulate' is doing heavy diplomatic lifting here. If you're making maximalist demands on day seven, you're either supremely confident in your position or you've already accepted this drags on for months.

China's 300bn yuan bank recapitalisation reveals a telling pattern: the biggest banks get bailed out whilst smaller ones - who actually have the worst bad loan problems - are left to consolidate.

Kim's observation rings true: AI has flattened the advantage that junior engineers had through pure algorithm grinding. The premium now shifts to senior engineers who understand system design, infrastructure trade-offs, and architectural patterns that no LeetCode prep teaches.

Why the Pentagon Wants to Destroy Anthropic | The Ezra Klein Show YouTube video by The Ezra Klein Show

Incredible episode by @ezraklein.bsky.social

Aside from Anthropic<->DoW, this thoroughly combs through ethics & politics. I honestly feel anyone could watch this episode and their views on AI would simply be up-leveled, not changed.

www.youtube.com/watch?v=xc97...

Current AI research evaluations use closed benchmarks that miss personalisation entirely. PDR-Bench creates 250 realistic scenarios pairing research tasks with authentic user profiles - finally testing whether AI agents can actually adapt to individual contexts rather than generic queries.

Xi's anti-welfarism ideology meets the AI employment cliff. Bloomberg flags the contradiction: mass AI layoffs could force Beijing to abandon its philosophical opposition to extensive social safety nets. Economic reality trumping ideological consistency.