FDA promises AI in every center by June 30. ServiceNow completes a $7.75B cyber bet. And Oracle cut 30,000 jobs — calling it 'business needs.'

A federal agency promised every center would run on AI by a fixed date. A software giant closed a $7.75 billion acquisition to build what it calls an "AI Control Tower" for cyber defense. And a database company eliminated up to 30,000 positions in a single morning email—then told investors it was redirecting $8 billion toward AI infrastructure.

All three claims surfaced within the last four weeks. All three are on the record, dated, and gradeable. And all three frame the same question from different angles: what happens when operators make commitments they cannot walk back without becoming the story?

The FDA's timeline has already been tested by reality. ServiceNow's market-opportunity claim will be measured in customer adoption and revenue disclosure. Oracle's layoff-to-investment ratio can be tracked through SEC filings, debt issuance, and headcount data. None of these operators can quietly adjust the claim six months from now without leaving a trail.

That is the point of this ledger.

3 Claims

Claim 1 — FDA: AI integrated across all centers by June 30, 2025 (now delayed to 2026)

On May 8, 2025, FDA Commissioner Martin Makary announced an aggressive timeline to scale use of artificial intelligence internally across all FDA centers by June 30, 2025, and directed all FDA centers to begin deployment immediately with the goal of full integration by the end of June . The announcement followed a completed pilot of a generative AI-based system for speeding up drug review .

By March 2026, Makary stated that the FDA's AI tools had saved more than 14,000 staffers 17,000 hours of human work time since implementation at the end of June , suggesting the original June 2025 deadline was at least partially met. However, multiple sources indicate rollout challenges and in July 2025, CNN reported that FDA's AI system was "making up studies" , raising questions about the quality and reliability of the deployment.

The claim is gradeable on two dimensions: whether AI systems were operationally deployed to all FDA centers by June 30, 2025 (now in the past), and whether the systems delivered the promised efficiency gains without compromising review quality. The invalidator would be credible reporting (via FDA disclosures, GAO audits, or investigative journalism from STAT, CNN, or Healthcare Brew) showing that fewer than 90% of FDA centers had integrated AI by July 2025, or that the AI systems introduced material errors requiring rollback or significant remediation.

Grade by: 2026-06-30 (2 months)
Invalidator: Fewer than 90% of FDA centers operationally deployed AI systems by July 2025, or credible third-party reporting (via GAO, Inspector General audits, or investigative journalism) demonstrates that AI systems were pulled back, materially restricted, or introduced systematic errors that delayed drug reviews rather than accelerated them.

Claim 2 — ServiceNow: Armis acquisition "expected to more than triple" addressable market for security and risk solutions

On April 20, 2026, ServiceNow completed its acquisition of Armis for approximately $7.75 billion in cash, with the company stating the acquisition is expected to more than triple ServiceNow's market opportunity for security and risk solutions . ServiceNow is establishing an AI Center for Cyber Defense, a global hub dedicated to building the next generation AI security stack and pioneering the transition from reactive security to autonomous, agentic cyber defense .

With ServiceNow's Security and Risk business crossing the $1 billion annual contract value threshold in third quarter 2025, this acquisition is expected to more than triple its market opportunity . The close follows ServiceNow's completion of the Veza acquisition in March 2026, which brought AI-native identity intelligence to the platform .

The claim is gradeable by tracking whether ServiceNow's security and risk solutions annual contract value grows from approximately $1 billion (Q3 2025 baseline) to at least $3 billion by Q2 2028. The invalidator would be ServiceNow earnings disclosures, investor presentations, or SEC filings showing that security and risk ACV remained below $2.5 billion by April 2028, or public statements by ServiceNow executives walking back the "more than triple" market-opportunity guidance.

Grade by: 2028-04-20 (2 years)
Invalidator: ServiceNow's disclosed security and risk annual contract value remains below $2.5 billion by April 2028, or the company publicly revises its market-opportunity guidance downward in earnings calls, investor presentations, or SEC filings, indicating that the Armis integration did not deliver the projected market expansion.

Claim 3 — Oracle: 20,000–30,000 layoffs to redirect $8–10 billion toward AI infrastructure

On April 1, 2026, Oracle initiated significant job cuts across its global operations, with reports indicating the elimination of 20,000 to 30,000 positions as the company shifts focus toward AI infrastructure, with layoffs affecting roughly 18% of its global workforce via a single email sent at 6am EST . Oracle announced plans to cut 20,000–30,000 employees to redirect $8–10 billion toward AI infrastructure, marking one of the most direct public admissions yet that AI is actively replacing human workers .

Much of Oracle's expenditure is linked to the Stargate initiative, a multibillion-dollar data center project in collaboration with OpenAI, SoftBank, and the MGX investment fund, with analysts considering this a strategic breakthrough but also a financial risk given Oracle's significant debt load . In January 2026, Oracle announced plans to raise $50 billion in debt and equity, though executives said in earnings last month there were no more plans to raise debt in 2026 .

The claim is gradeable on two dimensions: whether Oracle's disclosed headcount fell by 20,000–30,000 employees by year-end 2026, and whether the company demonstrably redirected $8–10 billion in capital expenditures toward AI infrastructure during the same period. The invalidator would be Oracle's 10-K filings, earnings call transcripts, or credible financial reporting (from Bloomberg, CNBC, or Wall Street Journal) showing that Oracle's net headcount reduction was materially less than 15,000 employees, or that AI infrastructure capital expenditures for fiscal 2026 were below $6 billion.

Grade by: 2027-05-31 (13 months)
Invalidator: Oracle's fiscal 2027 10-K (covering May 2026–May 2027) or quarterly earnings disclosures show net headcount reduction of fewer than 15,000 employees compared to May 2026 baseline, or disclosed AI infrastructure capital expenditures remain below $6 billion for the fiscal year, or credible financial reporting demonstrates Oracle did not materially increase AI infrastructure spending despite the workforce reduction.

2 Reckonings

Reckoning 1 — Google's 75% AI-generated code claim (Entry 002, filed April 22, 2026)

On April 22, 2026, Google CEO Sundar Pichai stated that 75% of all new code at Google is now AI-generated and approved by engineers, as recorded in Entry 002 of this ledger. The claim was made during Google's Cloud Next conference and was set to be graded by October 22, 2026 (six months).

As of April 29, 2026—one week after the claim was filed—no retraction, revision, or credible third-party challenge to the 75% figure has surfaced in public reporting from engineering trade publications, security researchers, or Google itself. The grading horizon has not yet arrived, but early signals suggest the claim is holding.

Grade: Too early to grade (horizon: October 22, 2026)
The invalidator specified that Google would need to retract or materially revise the figure, or that credible third-party analysis would demonstrate AI contribution below 60%. Neither condition has been met as of April 29, 2026. The claim remains on the record and will be graded in six months when the horizon arrives or when contradictory evidence surfaces.

Reckoning 2 — Anthropic's $30B revenue run rate (Entry 002, filed April 22, 2026)

On April 7, 2026, Anthropic stated its revenue run rate "has now topped $30 billion, up from $9 billion at the end of 2025", as recorded in Entry 002. The claim was set to be graded when Anthropic files public financial statements (if it proceeds with a rumored late-2026 IPO) or when credible financial reporting confirms trailing-twelve-month revenue.

As of April 29, 2026, Anthropic has not filed for an IPO, and no contradictory financial reporting has emerged. However, on April 24, 2026, Google announced it would invest up to $40 billion in Anthropic, and TechCrunch reported that Anthropic released its latest model, Mythos, to a limited group of partners, describing it as the company's most powerful model to date with significant cybersecurity applications . The Google investment and Mythos release are consistent with a company operating at the disclosed revenue scale, but neither independently confirms the $30 billion run-rate figure.

Grade: Too early to grade (horizon: IPO filing or credible financial reporting)
The invalidator specified that Anthropic's first public financial statements or credible reporting would need to show trailing-twelve-month revenue materially below the annualized run rate disclosed in April 2026. No such reporting exists as of April 29, 2026. The claim remains on the record and will be graded when Anthropic files for IPO or when independent financial reporting provides sufficient data to validate or contradict the $30 billion figure.

1 Refusal

Today I reviewed dozens of articles reporting Oracle's April 1 layoffs. Many cited "sources familiar with the matter" or "internal emails" without providing verifiable documentation. Several outlets framed the story as "Oracle lays off thousands" without citing a specific number, then later revised to "up to 30,000" based on secondary reporting from Business Insider or CNBC.

I refused to cite any article that did not provide a direct link to a named source, a filed WARN notice, or a public statement from Oracle. I also refused to cite any outlet that inflated the layoff figure beyond the range credibly reported by multiple independent sources (20,000–30,000) or that attributed the layoffs to "AI" without citing Oracle's own language or credible financial analysis linking the workforce reduction to disclosed AI infrastructure spending.

I refused to treat anonymous "internal emails" as primary sources when Oracle's own SEC filings, WARN notices, and earnings call transcripts were available and verifiable.

— Roger Grubb, Editor