"Every company is going to have to build what I think of as human capital and token capital. Human capital comprises the knowledge, judgment, relationships, ingenuity, and pattern recognition of its people, while token capital is the firm’s AI capability it builds and owns."

Simon Willison and Arvind Narayanan published analyses on June 11 and 14, 2026, arguing that the structural reasons software engineers have not been displaced by AI are durable rather than temporary. The argument rests on a decide-execute-deliver framework: AI has compressed the execution layer of software development, but the decision layer, which requires understanding user needs and organisational priorities, and the delivery layer, which requires accountability for production systems, resist automation for structural reasons. A Harvard Business Review survey found that 21% of executives had made large headcount reductions in anticipation of AI, while only 2% had done so due to actual AI implementation. Full details via Normal Tech / normaltech.ai/p/why-ai-hasnt…

France's Choose France summit on June 1, 2026, produced announced AI investment commitments of €93bn, with SoftBank's pledge of up to €75bn for 5 gigawatts of AI data centre capacity in northern France accounting for the majority. The EU's parallel plan for 5 AI gigafactory data centres has been delayed, with bidding pushed from May to July 2026. Only 2 of the 5 centres are fundable before the EU's next budget cycle begins in 2028, and the number of expected bidders has narrowed from approximately 70 companies to approximately 10. The Anthropic export control episode broke the weekend before the G7 summit in Évian, complicating the transatlantic AI investment narrative French President Emmanuel Macron had been constructing. Full details via France 24 / france24.com/en/france/20260…

Canadian Prime Minister Mark Carney, speaking in Westport, Ireland on June 14, 2026, used the Anthropic export control action to argue publicly for sovereign AI infrastructure and multi-provider architectures. Carney drew an explicit parallel with the 2008 financial crisis, framing AI model concentration as a systemic risk analogous to the interconnected bank dependencies that amplified that crisis. He called for "redundancy and diversity" in AI development ahead of the G7 summit in Évian-les-Bains, where AI was expected to dominate the agenda. A European Commission spokesman separately confirmed the bloc was reviewing the situation and that the episode had reinforced calls within Europe for greater independence from foreign technology suppliers. Full details via Quartz / qz.com/carney-us-ai-restrict…

📰 The Daily brief / Monday June 15, 2026 / Anthropic enters crisis management after US export control order disables Fable 5 and Mythos 5 globally, affecting banks, government agencies, and an IPO targeting $1tn / Amazon chief executive Andy Jassy personally contacted Treasury Secretary Scott Bessent to trigger the federal action that shut down Anthropic's two most advanced models within 90 minutes / Canadian Prime Minister Mark Carney invokes the 2008 financial crisis to argue that AI model concentration is a systemic risk, speaking ahead of the G7 summit in Évian / Macron's Choose France summit produced €93bn in announced AI investment commitments, but the EU gigafactory programme is delayed and only 2 of 5 centres are fundable before 2028 / Bloomberg documents a 77% decline in London finance analyst job postings over 4 years, with junior roles in legal services and technology contracting alongside financial services / Simon Willison and Arvind Narayanan argue that structural reasons software engineers have not been displaced by AI are durable, grounding the case in a decide-execute-deliver framework / Google Cloud publishes the Open Knowledge Format, a vendor-neutral Markdown standard designed to make organisational knowledge traversable by AI agents 🤖 made with AI ✅ reviewed by the Tokenando team 🔗 links in comments

Microsoft CEO Satya Nadella, speaking at a live taping of The New York Times' "Hard Fork" podcast in mid-June 2026, acknowledged that "a lot" of tokenmaxxing is happening inside Microsoft. "I'm a tokenmaxxer too, it's addictive. But you have to step back when the novelty wears off to say, 'What is it that I'm trying to create?'" Mr Nadella directed employees toward Copilot's auto mode, which selects the model best suited to each query, with the principle: "Don't use frontier models for non-frontier problems." The comments came within days of Microsoft's announcement that all employee Claude Code licences would be terminated on June 30, 2026, coinciding with the company's fiscal year end. Microsoft reports that 30% of its code is now written by AI. Full details via The Decoder / the-decoder.com/microsoft-ce…

📰 The Daily brief / Sunday June 14, 2026 / AI ventures absorb $297 billion in Q1 2026, with 4 companies capturing 65% of all global venture capital in a single quarter / Meta warns 6,000 employees of billions in AI costs after internal token leaderboard drives 73.7 trillion tokens in 30 days / Kimi K2.7 Code launches at $4.00 per million output tokens, undercutting Claude Fable 5 by more than 12 times on price / Microsoft CEO Satya Nadella calls tokenmaxxing "addictive" days before terminating all employee Claude Code licences on June 30 / Tsinghua University's Count Anything model achieves a mean absolute error of 9.34 objects per image across 6 domains, against 23.38 for the next best method 🤖 made with AI ✅ reviewed by the Tokenando team 🔗 links in comments

The US Department of Commerce ordered Anthropic on June 13, 2026 to suspend access to its Fable 5 and Mythos 5 models for all foreign nationals, whether inside or outside the United States, including Anthropic's own non-US employees. Anthropic received the directive at 5:21pm ET and disabled both models globally, concluding it could not selectively enforce the restriction. The government's stated basis was a jailbreak of Fable 5 involving prompts that identify software vulnerabilities in codebases. Anthropic disputed the severity, noting the same capability exists in other publicly available models including OpenAI's GPT-5.5. Fable 5 had been publicly available for 4 days. Mythos 5 had been accessible to roughly 50 vetted partners under Project Glasswing, including Amazon, Apple, Google, Microsoft, and CrowdStrike. The suspension follows the Trump administration's earlier designation of Anthropic as a "supply chain risk" after the company refused to allow its models to be used for domestic surveillance and fully autonomous weapons systems. Full details via Anthropic / anthropic.com/news/fable-myt…

📰 The Daily brief / Saturday June 13, 2026 / US Department of Commerce orders Anthropic to suspend Fable 5 and Mythos 5 globally, citing a disputed jailbreak finding 4 days after public release / Anthropic's Claude Fable 5 costs twice as much as its predecessor for a 5.7% aggregate performance gain, with a hidden fallback billing issue affecting up to 8% of queries / A coalition of US state attorneys general, led by New York, serves OpenAI with a formal subpoena covering advertising, data practices, and child safety, days after the company files confidentially for a $1 trillion IPO / SpaceX rented its Colossus 1 data centre to Anthropic after its own teams could not operate the facility effectively, with GPU utilisation running at 11% against an industry standard of 35 to 45% / Mistral AI enters funding discussions targeting €3bn at a €20bn valuation, nearly doubling its September 2025 Series C figure, as it builds out owned compute infrastructure in France and Sweden / OpenAI introduces rate-limit banking for its Codex coding agent, letting users save and deploy resets on demand, as the company weighs token price cuts in response to Anthropic's Fable 5 launch / Google files a civil lawsuit with FBI support against a Chinese phishing network that used Gemini to build fraudulent sites, as OpenAI separately discloses it banned 2 clusters of PRC-linked ChatGPT accounts 🤖 made with AI   ✅ reviewed by the Tokenando team   🔗 links in comments

The European Commission published the Chips Act 2.0 on June 3 as part of a broader European Technological Sovereignty Package that also includes the Cloud and AI Development Act and an EU Open Source Strategy. The Chips Act 2.0 proposes accelerated permitting with a maximum 12-month approval timeline and support for strategic projects including a potential advanced semiconductor facility. Reports indicate the Commission is working with Belgian R&D centre Imec and German innovation agency SPRIND on plans for a facility capable of producing chips at leading-edge nodes, with one option being a one-nanometre factory in Germany by 2035. The initiative could cost up to €50bn. The global semiconductor market is projected to reach €1.37tn by 2030, with AI-related components accounting for roughly 70% of that growth. The legislation must still be negotiated with EU member states and the European Parliament before entering into force. Full details via CEPA / cepa.org/article/europe-purs…

The Financial Times published an opinion piece on June 11 arguing that existing tax frameworks are structurally unsuited to capturing the economic value generated by AI, particularly where productivity gains accrue to capital rather than labour and where the geographic location of value creation is ambiguous. The piece is the most prominent mainstream business media treatment of the question in the current cycle. A Brookings Institution paper in January set out a public finance framework for the AI age; a Time piece in May called for an excise tax on data centre energy use; a Pictet Asset Management analysis in June described wealth taxes as "almost inevitable" given the concentration of AI-driven gains. Instruments under active discussion include token taxes, robot taxes, consumption tax shifts, and digital services tax extensions to cover AI and cloud. Full details via Financial Times / ft.com/content/b277360e-bf23…

📰 The Daily brief / Thursday June 11, 2026 / Oracle posts record $638bn backlog and 93% cloud growth but capex of $55.7bn overshoots guidance, sending shares down roughly 7% to 9% in after-hours trading / OpenAI bans 2 ChatGPT account clusters linked to China after they generated content designed to amplify US data centre opposition / Anthropic reverses a covert Claude Fable 5 restriction that silently degraded outputs for AI research queries, following public backlash from researchers / Anthropic security research shows Claude Mythos Preview built working Windows kernel exploits in under 6 hours at roughly $2,000 per exploit chain / Financial Times publishes substantive argument that existing tax frameworks cannot capture AI's economic value, as Brookings, Pictet, and others add to the policy debate / European Commission publishes Chips Act 2.0 with a potential €50bn sovereign AI chip facility and a one-nanometre factory in Germany targeted for 2035 🤖 made with AI ✅ reviewed by the Tokenando team 🔗 links in comments

On June 10, 2026, Ben Thompson published an analysis of the Claude Fable 5 launch on Stratechery, examining the commercial architecture of Anthropic's Mythos tier. Thompson argues the Fable/Mythos split reflects a deliberate attempt to segment the market by willingness to pay and by organisational trust level. The Glasswing membership list includes NATO, Samsung, SK Hynix, Okta, and ENISA, all US-aligned entities. Thompson contrasts Anthropic's approach with OpenAI's operator-tier model, which makes capability broadly available and differentiates on customisation rather than on access to the underlying model. Full details via Stratechery / stratechery.com/2026/fable-5…

📰 The Daily brief / Wednesday June 10, 2026 / Anthropic releases Claude Fable 5 for general use and Mythos 5 for vetted partners, Fable is priced at double Opus 4.8 pricing, with safety classifiers routing up to 5% of queries to the cheaper model / Google revealed as payment backstop across all 5 data centres in Apollo and Blackstone's $35bn chip financing deal for Anthropic / Munich Regional Court issues Germany's first injunction holding Google directly liable for false claims generated by its AI Overview feature / Stratechery's Ben Thompson argues Anthropic's Mythos tier is a pricing segmentation strategy, with the Glasswing access list reading as a geopolitical alignment map / Simon Willison spends $110 in a single Fable 5 session completing several days of engineering work, then flags a system card disclosure on silent output modification for AI development requests 🤖 made with AI ✅ reviewed by the Tokenando team 🔗 links in comments

Agentic AI coding tasks consume approximately 1,000 times more tokens than equivalent code-chat tasks, according to research from Stanford, Michigan, DeepMind, and MIT published in analysis by The Decoder on June 8, 2026. The same research found that on identical tasks, different frontier models consumed up to 1.5 million more tokens than each other, and that 4 runs of the same model on the same task can produce up to 30 times variance in total token cost. Mercor, an AI-native startup operating at over $1 billion in annualised revenue, reported in early June that its internal token spend on AI agents now exceeds its employee salary costs. Full details via The Decoder / the-decoder.com/frontier-rad…

📰 The Daily brief / Tuesday June 9, 2026 / OpenAI files confidentially with the SEC for an IPO, joining Anthropic in a dual listing race that could bring close to $3 trillion in new AI market capitalisation to public markets within months / Apple launches Siri AI with Google Gemini at its core under a reported $1 billion annual licence, then confirms the feature will not reach EU iPhones or iPads at launch due to a DMA compliance deadlock / Google places a firm order with Intel for more than 3 million tensor processing units in 2028, while Nvidia separately evaluates Intel's 18A process as a potential TSMC alternative / ASML chief executive Christophe Fouquet warns the European Commission against directing strategic chip projects, noting that 99% of ASML's machine sales already go outside Europe / Electrical contractors, power equipment makers, and cooling specialists are reporting doubled margins and multi-year backlogs as hyperscalers accelerate data centre construction / Stanford and MIT research finds agentic AI coding tasks consume approximately 1,000 times more tokens than equivalent chat tasks, as Mercor reports its token spend now exceeds its employee salary bill 🤖 made with AI ✅ reviewed by the Tokenando team 🔗 links in comments

Elon Musk is scheduled to participate in a virtual fireside chat with ASML chief executive Christophe Fouquet at ASML's annual internal technology conference in the week of June 12, 2026. The event is not open to the public. According to internal ASML communications, Musk will discuss AI, robotics, space, and semiconductor manufacturing, with the Terafab concept as a central topic. Terafab is a joint venture between SpaceX, Tesla, and Intel to build a network of large-scale semiconductor manufacturing facilities in Texas. SpaceX has filed for a $55bn facility in Grimes County; total project costs could reach $119bn. Intel has committed its 14A process technology to the project. Fouquet confirmed in May 2026 that he had spoken directly with Musk and described him as "very serious" about the project. ASML is the sole global supplier of extreme ultraviolet lithography machines required for leading-edge chip manufacturing. Full details via Business Insider / africa.businessinsider.com/n…

Tokyo-based Sakana AI has formally established the Sakana AI RSI Lab, a dedicated research group developing recursive self-improvement systems: AI that iteratively redesigns its own architecture, code, and training processes. The lab is based in Tokyo and is actively recruiting frontier research scientists. The announcement builds on prior Sakana research including the Darwin-Gödel Machine, which improved its own performance on the SWE-bench coding benchmark from 20% to 50% through iterative self-modification, and the RL Conductor, a 7-billion-parameter model that outperformed GPT-5 on graduate-level science and competition mathematics benchmarks by coordinating multiple frontier models. Sakana's stated thesis is that recursive self-improvement is achievable on modest compute rather than the hyperscale GPU clusters that define current frontier lab competition. Full details via The Decoder / the-decoder.com/sakana-ai-be…