Business, Deals & Funding
Guardian AI
Australia news live: SA premier announces royal commission into AI; Icac witness questioned on fake invoices says ‘every tradie in Sydney has wife as the director’
This Australian news live blog covers several stories: South Australia's premier has announced a royal commission into artificial intelligence; an ICAC (Independent Commission Against Corruption) witness being questioned about fake invoices defended the practice by claiming 'every tradie in Sydney has wife as the director'; NSW police confirmed that remains found in a suitcase near Goulburn are not human; and there are mentions of various other ongoing stories including transport, AUKUS public inquiry, housing undersupply, and the Alan Jones trial.
Why it matters
This is a typical Australian live news blog aggregating multiple developing stories. The AI royal commission in South Australia is noteworthy as it signals growing governmental concern about artificial intelligence regulation. The ICAC witness's casual dismissal of fake invoicing practices as commonplace is both darkly humorous and revealing about potential systemic corruption in Sydney's construction industry. The suitcase remains story provides a relieving resolution. The format, while useful…
Lenny's Newsletter
The playbook for building high talent density teams | Adam Ward, Head of Talent at Cursor
This podcast episode features Adam Ward, Head of Talent at Cursor, discussing strategies for building high-talent-density teams. Key topics include the current bifurcated talent market, why the traditional recruiting funnel (which he calls the 'funnel of doom') leads to mediocre hires, and his three-step playbook of scoping, mapping, and relentless pursuit. Ward also covers the rise of the forward deployed engineer role, the worst question to ask when sourcing great talent, the biggest mistake founders make when hiring their first recruiter, and why genuine caring is the biggest unfair advantage in recruiting. Ward brings over 20 years of experience building elite teams and previously founded Growth by Design, a recruiting and talent strategy firm serving ambitious AI and technology companies.
Why it matters
Based on the episode description and topics covered, this appears to be a highly valuable conversation for founders, hiring managers, and anyone involved in building teams at fast-growing companies. The concept of the 'funnel of doom' is a compelling reframe of conventional recruiting wisdom, and the idea of treating every hire like an executive search suggests a level of rigor that most companies lack. Adam Ward's position at Cursor—one of the most talked-about developer tools—gives him unique…
TechCrunch AI
Embattled hedge fund Situational Awareness invests $400M in chip startup Source Foundry
Leopold Aschenbrenner's AI-focused hedge fund Situational Awareness invested $400 million in Source Foundry, a Stanford-founded chip manufacturing startup, bringing its total investment to $500 million. The fund has faced significant challenges recently, having sold off most of its public portfolio to Citadel at the end of July amid declining AI infrastructure stocks, with assets under management dropping from $20 billion to $10 billion. Aschenbrenner, a former OpenAI researcher in his mid-twenties with no prior trading experience, launched the fund in 2024 and saw strong early returns before the recent downturn.
Why it matters
This article paints a fascinating picture of the volatility in AI investing and the risks of a young, inexperienced fund manager making enormous concentrated bets. Losing half of AUM ($20B to $10B) is a staggering decline, yet doubling down with a $400M investment in a chip startup while the portfolio is hemorrhaging shows either remarkable conviction or reckless overconfidence. The fact that Aschenbrenner went from writing influential AI essays to managing billions with no trading background m…
The Verge AI
AI detectors are creating a new era of distrust
The Verge article by Emma Roth discusses how AI writing detection tools are fostering a new climate of distrust in educational and publishing settings. While anti-plagiarism tools like Turnitin existed before ChatGPT, the rise of generative AI has led to widespread adoption of AI detection tools that claim to identify machine-generated text. However, these tools can be unreliable and dubious in their accuracy, yet educators and publishers continue to use them anyway, creating problems for writers and students who may be falsely accused of using AI to produce their work.
Why it matters
This article highlights a genuinely important and troubling development in the AI era. AI detection tools are fundamentally flawed because they attempt to solve a problem that may be technically unsolvable — reliably distinguishing human writing from AI-generated text. The false positive problem is particularly concerning, as innocent students and writers can face serious academic or professional consequences based on unreliable algorithmic judgments. The rush to adopt these tools reflects a br…
Guardian AI
Meme factories are churning out hard right AI slop on Facebook | First Dog on the Moon
This article features a First Dog on the Moon cartoon series published in The Guardian, highlighting how meme factories are producing AI-generated pro-One Nation (an Australian far-right political party led by Pauline Hanson) content, referred to as 'slop,' on Facebook. The cartoonist expresses alarm at the proliferation of hard-right AI-generated political propaganda on social media platforms.
Why it matters
The article, presented as a political cartoon series, takes a clearly critical and alarmed stance toward the use of AI-generated content to promote far-right political messaging on Facebook. The cartoonist views the phenomenon of AI 'slop' supporting One Nation as frightening and dangerous, positioning it as a threat to political discourse. The tone is satirical but genuinely concerned about the intersection of AI content generation and far-right political propaganda on social media platforms.
TechCrunch AI
Anthropic is turning Claude Code’s auto mode on by default
Anthropic is making Claude Code's auto mode the default for Pro, Max, and Team accounts starting August 14, 2026. Auto mode allows Claude Code to proceed with actions without human approval at each step, unless an action is deemed irreversible, destructive, or aimed outside the user's environment. Anthropic cites a study of 1,053 paid testers showing auto mode caught 89% of harmful actions compared to only 13.6% caught by human review, noting that users approve 97% of permission prompts manually anyway. The company has also added safety features like prompt injection screening and customizable hard deny rules to prevent data exfiltration.
Why it matters
This is a significant and somewhat concerning development in AI-assisted programming. While the statistics about human review being largely rubber-stamped (97% approval rate) are compelling, making reduced oversight the default rather than opt-in raises important questions. The 89% catch rate for harmful actions, while better than human review, still means 11% of harmful actions slip through automatically — and without a human in the loop by default, those failures could compound. The framing t…
Guardian AI
UK manufacturers face rising hacking risk as survey shows 30% were hit last year
A survey reveals that nearly a third of UK manufacturers have experienced a cyber-attack on themselves or their supply chain in the past year. The findings highlight the growing hacking risk to companies, coming almost a year after JLR, Britain's largest automotive employer, suffered an attack that halted production for weeks. Despite being under constant threat, only half of large companies have a response plan in place.
Why it matters
This article underscores a critical and growing vulnerability in the UK manufacturing sector. The fact that 30% of manufacturers have been hit by cyber-attacks while only half have response plans is alarming and suggests a significant gap between threat awareness and preparedness. The JLR incident serves as a stark reminder of the real-world consequences of inadequate cybersecurity, including production shutdowns and supply chain disruptions. The manufacturing sector's increasing digitization a…
TechCrunch AI
Historian Jill Lepore says Silicon Valley misreads science fiction and undermines democracy
Historian Jill Lepore discusses her upcoming book 'The Rise and Fall of the Artificial State' on TechCrunch's Equity podcast, arguing that tech companies are increasingly replacing democratic government functions, constituting 'a return to tyranny and mystification in the form of rule by algorithms, corporations, machines.' She contends that Silicon Valley leaders, especially Elon Musk, are guided by misread science fiction. Lepore clarifies she is not anti-technology but opposes private corporations usurping state functions without public consent, a shift she says has been gradual and largely accidental but has become more deliberate in the last 20-25 years.
Why it matters
This is a thought-provoking interview that raises important questions about the growing power of tech corporations over functions traditionally belonging to democratic governments. Lepore's framing of the 'artificial state' is compelling and her distinction between being anti-technology and being critical of corporate overreach is an important nuance often lost in these debates. The observation that tech leaders misread the very science fiction they claim to be inspired by is a sharp and entert…
From X/Twitter
- Garry Tan spent 42 minutes at Startup School 2026 making the case that the leverage is in your context, not the model.
- Boris Cherny breaks down how prompt injection remains the easiest way scammers hijack AI agents — one malicious webpage and your SSH keys are gone.
- Built a site in 3 minutes for 47 credits using opencode and Netlify Agent Runners — token efficiency is the new flex.
- HuggingFace's speech-to-speech framework (11.8K+ GitHub stars) lets you run fully local voice AI agents with zero cloud API costs.
From Reddit/HN/YC
- [Hacker News] Vincent Schmalbach says switching from GPT-5.5 to GPT-5.6 made him less productive — bigger model, worse workflow.
- [Hacker News] PrimeIntellect introduces Recursive Language Models and calls them the paradigm of 2026, with an open-source self-improving agent to match.
- [Hacker News] Flyn is a link shortener that counts bots separately from humans — finally.
- [Hacker News] TBD is a native, minimalist macOS front end for yt-dlp that does exactly what you'd hope.
- [Hacker News] Harry Potter fans force the reroute of a £430M undersea cable to avoid Dobby's sandy grave.
- [Hacker News] South Australia launches a royal commission into artificial intelligence — a first of its kind.
- [Hacker News] Remembering the pre-Google web, when search was an experiment and nobody had won yet.
- [Hacker News] PiShot: a custom open-source Raspberry Pi camera with a modular 3D-printed body.
- [Hacker News] Hetzner quietly launches an experiments platform with an inference API — cheap GPU hosting gets more interesting.