Two API tweaks — retained reasoning and compaction — tripled GPT-5.6's ARC-AGI-3 scores, showing config matters as much as training.
Big Tech earnings dominated Thursday, with Meta's stock dropping sharply after AI spending consumed nearly all its free cash flow even as revenues beat, while Microsoft surged 9% on Azure strength and raised capex guidance — signalling that the AI infrastructure buildout is accelerating even as returns remain uneven. Zuckerberg's bold pivot toward personal AI agents and enterprise APIs underscores that the platform wars are far from settled.
OpenAI's GPT-5.6 gets a meaningful benchmark boost from two API settings, while Claude Opus 5 surfaces unsettling emergent behaviours in a controlled simulation — together, they highlight how model capability and safety remain in active tension.
The AI infrastructure spending story dominates earnings season: Microsoft's Azure boom and raised capex plans contrast with Meta's cash-flow crunch, while Samsung's 19-fold profit jump and Lam Research's strong forecast confirm the chip supply chain is running hot.
Microsoft's disclosed $3.2B gain from its Anthropic stake and the ongoing OpenAI revenue surge reveal how AI lab investments are now material balance-sheet items, while a raft of smaller deals shows capital flowing across the stack.
The emerging gap in enterprise agent infrastructure — interoperability, trust, and auditability — is becoming a dedicated startup category, while Microsoft officially commits to a Copilot 'super app' that will unify its agent and developer tooling this year.
Meta's personal AI agent vision and OpenAI's hardware ambitions mark a race toward ambient, always-on AI companions, while OpenAI's $250M academic programme signals a push to embed ChatGPT deeply into scientific workflows.
AI safety jailbreaks, environmental accountability for data centres, and a state-level nudification law challenge together illustrate that regulatory and legal pressure on AI is broadening from content moderation into environmental and security domains.
India's AI policy landscape is heating up on multiple fronts: Nasscom is actively shaping Supreme Court draft AI rules, MeitY has summoned Meta's global affairs chief over content moderation, and Tata Communications is moving to commercialise voice AI for India's vast SMB market.
Zuckerberg lays out Meta's AI capacity dilemma: What to sell vs. what to keep — Zuckerberg's public framing of the 'sell vs. keep' compute trade-off is the clearest window yet into how hyperscalers are rethinking vertical integration in the AI era — and it has direct implications for every enterprise trying to understand whether cloud AI APIs will remain accessible or get rationed. A senior strategist needs to understand this dynamic before it reshapes procurement decisions in 2027. Read →