Oracle fired 30,000 with a 6am email. OpenAI wants robot taxes and a four-day workweek. Microsoft broke from OpenAI with 3 new models. Gartner declared copilots dead. Eight stories from the week agentic AI started billing for its externalities.
Oracle fired 30,000 people with a single 6am email on March 31 — the largest layoff in the company's 47-year history, executed without warning. That's the opening act. This week, OpenAI published a 13-page economic manifesto calling for robot taxes, public wealth funds, and a four-day workweek — a document that reads as both confession and campaign platform from a company valued at $852 billion. Meanwhile, Microsoft declared independence with three in-house AI models, Google shipped Gemma 4 under Apache 2.0, and OpenAI's C-suite lost its COO, AGI CEO, and CMO in five days flat. Eight stories from the week agentic AI started billing for its externalities.
Oracle Sent 30,000 Termination Emails at 6am.
The Biggest Layoff in Its 47-Year History.
On the morning of March 31, thousands of Oracle employees around the world opened an email to find out they no longer had jobs. No warning from HR. No conversation with a manager. Just a message from "Oracle Leadership" confirming their role had been eliminated, effective immediately — with US employees' last official day set for April 10. Investment bank TD Cowen estimated the cuts at between
of Oracle's global workforce of 162,000. The stated purpose: freeing up
AI data center buildout. This is Oracle's largest-ever layoff by a significant margin, and the most explicit corporate acknowledgment yet that AI infrastructure investment requires human capital to be liquidated at scale.
Oracle's layoff memo is a case study in what the agentic transition looks like from the inside: people out, GPUs in, no apology. The 6am email is a detail that will age badly — companies that treat workforce transitions as security events rather than human moments will spend years rebuilding trust. The AI buildout is right. The execution is a cautionary tale.
OpenAI Wants Robot Taxes, a Public Wealth Fund,
company is calling for taxes on itself. On April 6, OpenAI released a 13-page policy blueprint titled "Industrial Policy for the Intelligence Age: Ideas to Keep People First," proposing a nationally managed public wealth fund seeded by AI company contributions, a tax on robot-driven capital replacing labor income, and government-backed experiments with 32-hour workweeks at full pay. The parallel to Alaska's Permanent Fund is intentional — every American receiving an ownership stake in AI gains. The document arrives days after OpenAI COO Brad Lightcap stepped back from operations and with an IPO on the horizon. Reading it as pure altruism would be naive. Reading it as clever pre-IPO positioning and policy hedging is more accurate. But the policy ideas themselves are more serious than any competitor has produced.
The most important sentence in the document isn't about policy — it's the implicit admission that AI-driven productivity won't automatically distribute wealth without deliberate intervention. A company at $852B valuation calling for taxes on itself is either genuine or brilliant PR. Either way, it just moved the Overton window on AI policy further than any regulator has managed in three years.
Microsoft Built Three Frontier Models Without OpenAI.
On April 2, Microsoft launched three foundational AI models built entirely in-house:
(speech-to-text, 2.5× faster than Azure's existing offering, top-ranked across 25 languages),
(voice generation producing 60 seconds of audio per second, with emotional range and speaker identity preservation), and
(a top-3 image generator on the Arena.ai leaderboard). All three are available immediately through Microsoft Foundry. The timing is deliberate: as OpenAI moves toward IPO and potentially renegotiates its partnership terms, Microsoft is building the internal capability to operate without exclusive dependency. This is a hedge that took years to build — and this week it became public.
Microsoft just publicly revealed its OpenAI exit ramp. These three models span the highest-value enterprise AI modalities — speech, voice, and image — and they were built quietly, inside Microsoft, without fanfare. The partnership with OpenAI continues, but the negotiating dynamic just shifted. Quietly but permanently.
Google Drops Gemma 4 Open Source Under Apache 2.0.
Google DeepMind released Gemma 4 on April 2 under a fully permissive