Anthropic's reasoning model Mythos is here, Sam Altman reversed course on the AI jobs apocalypse, and Groq is raising $650M after Nvidia's $20B power move.
Anthropic launches "Mythos", and warns AI could self-improve by 2028
Anthropic released Mythos, an advanced reasoning model built for complex strategic planning tasks. On the same day, the company's Head of Policy stated there is a 60% probability that AI systems could begin recursively self-improving (essentially getting smarter on their own without human direction) by 2028. That's two years away.
Why it matters: Launching a powerful model and simultaneously flagging a near-term safety risk is unusual and deliberate. Anthropic is signalling that it takes its own warnings seriously, which is the kind of transparency that influences how other companies and regulators respond.
Sam Altman: "I was wrong about the jobs apocalypse"
OpenAI CEO Sam Altman publicly said AI is unlikely to trigger a global jobs apocalypse, admitting he had previously overestimated how quickly AI would eliminate white-collar jobs. He maintained that disruption will happen but argued the timeline is slower and more uneven than feared, with many professions adapting rather than disappearing.
Why it matters: When the person building the most powerful AI tools revises his view on job losses, it is worth noticing, especially for students and mid-career professionals deciding how much urgency to apply to AI skill-building. The answer seems to be: real urgency, but not panic.
Groq raising $650M as the AI chip race intensifies
AI chip startup Groq is reportedly raising $650M from existing investors, a day after Nvidia completed a $20B "not-acqui-hire" deal that reshaped the AI hardware landscape. Groq builds its own chips specifically optimised for running AI inference (the part of AI that generates answers) faster and cheaper than standard GPU setups.
Why it matters: The battle for AI infrastructure is now about who owns the chips, not just who trains the models. Faster, cheaper inference means AI products that cost less to run and respond quicker, that trickles down to every app you use.
Deepfake job candidates: 1 in 3 applicant profiles found to be AI-fabricated
Security firm Pindrop analysed 300 job applicant profiles and found that over one-third were entirely fabricated: AI-generated resumes paired with deepfake video interviews capable of passing live HR screenings in real time. Experian's 2026 Fraud Forecast names this as a top emerging threat, and Gartner projects 1 in 4 global candidate profiles will be fake by 2028.
Why it matters: This is not a future problem. If you're hiring, your screening process needs to account for this now. If you're job-seeking, being a verified, community-visible human is already a competitive edge.
Google is training all 6 million US teachers on AI, for free
Google launched the Google AI Educator Series, a free, comprehensive AI training programme available to all 6 million K-12 and higher education teachers in the United States, developed with ISTE+ASCD. New modules are added monthly. Google is investing over $150 million in AI literacy initiatives, including expanded access to NotebookLM for school and university accounts.
Why it matters: The quality of AI education your children receive in the next two to three years depends directly on whether their teachers have been trained. A programme at this scale moves that needle.
India's IndiaAI Mission scales to 40,000 GPUs; Gnani AI closes $10M Series B
India's IndiaAI Mission has expanded its national GPU pool from 18,417 to nearly 40,000 units, with further allocations expected in the coming weeks. On the startup front, voice AI company Gnani AI closed a $10M Series B round, adding to a string of Indian AI funding deals in recent weeks. The broader IndiaAI program has generated over $200 billion in committed investments following the India AI Impact Summit.
Why it matters: National compute access is the unglamorous foundation that makes everything else possible. India is building that foundation at speed, and startups like Gnani are beginning to build on top of it.
The 2028 self-improvement cliff, and why it changes everything
Anthropic's Head of Policy put a 60% probability on AI systems beginning to recursively self-improve by 2028. That phrase (recursive self-improvement) means AI that makes itself smarter without human guidance, compounding gains exponentially. If even 30% likely, it changes the calculus for policymakers, educators, and everyone building careers that depend on AI remaining a tool humans direct. Watch what other labs say in response. Silence will be as telling as agreement.
Boston makes AI a graduation requirement + OpenAI near IPO + EU deepfake labels due in August
Boston Public Schools will become the first major US city district to require AI fluency for graduation, starting September 2026. Separately, OpenAI has surpassed $25B in annualised revenue and is reportedly taking early steps toward a public listing, potentially before the end of 2026. And in Europe, the EU AI Act's requirement for AI-generated content to carry machine-readable labels takes effect August 2, 2026, about 60 days away.