Daily Brief Oct 31, 2026
How we checked the items below
For readers who want to dig in. What happened, in brief, is on the Today tab.
This issue draws on material from October 30 to early October 31; the From the archive item comes from SecondSource's October 23 deep dive. Last night's collection took in 432 pieces, and this issue uses 11 outside sources with checkable links; the later sections also draw on material from the past week.
1. [Today] Broadcom's second AI rack-lease backstop covers up to 72.5% of unpaid rent, capped at about US$43 billion, down from 85% in the first
Not proven yet： the customer, the US$63 billion and who lends what come only from 24/7 Wall St.'s summary
How we checked
The 72.5%, the US$43 billion and the payout formula come from the text of Broadcom's 8-K, a primary statutory filing. The 85%, the US$29 billion and the US$42 billion note facility come from the text of Broadcom's September 10 10-Q. The first deal's customer was named as Anthropic in a June 9 joint press release from Apollo, Blackstone and Broadcom (Apollo press release, 2026-06-09). Broadcom's 10-Q describes such guarantees as having "a low probability of occurring"; that is the guarantor's own assessment.
2. [Today] Oracle and Google say Google's TPUs come to Oracle's cloud from 2027, which the two call the first large cloud other than Google's to offer them
Not proven yet： scale, pricing and first customers haven't been announced
How we checked
The quote and timeline come from the text of the Oracle and Google joint press release, and the Arm sentence from the text of Arm's press release; both are the parties' own accounts. The scope of "first" is defined by Oracle and Google.
[This week] (preprint, arXiv 2610.38799) When several coding agents work together, a compromised testing or deployment component that hides malicious instructions in its status reports got Codex, OpenAI's coding agent, to act on them 61.3% of the time in the authors' tests.
How we checked
In the authors' self-reported tests, agents that treat other components' reports as trusted debugging guidance carry out actions nobody authorized. No other team has repeated these success rates yet, and the abstract doesn't say how many times each commercial product was tested.
[This week] (preprint, arXiv 2610.38521) Researchers put malicious instructions into CLAUDE.md, the project memory file Claude Code reads when it starts work; against runs without them, the share of secure code the agents wrote moved between -8 and +6 percentage points, with no consistent direction.
How we checked
In the authors' self-reported small sample, the clearer difference the paper measured is whether the agent tells the developer that the file influenced it.
[Today] OpenAI committed to honoring the preferences websites state in robots.txt on whether their content may be used to train AI; the labels come from a draft by the AI Preferences working group of the IETF, the body that writes internet protocol standards, and OpenAI expects to implement it within a few months.
How we checked
[Today] Google Cloud spend caps are now generally available: spending on AI services is counted from real-time cost estimates, and new requests are blocked within minutes of reaching the cap; requests on capacity reserved in advance are not blocked.
How we checked
[Looking back] (SecondSource deep dive, October 23) If AI data centers are overbuilt, long-term data-center leases and the debt of companies that rent out compute break before the big cloud companies' balance sheets do.
How we checked
Sources & accounting
The past 24 hours. 432 new pieces came in from October 30 to 31: 207 academic papers, 126 social-platform posts, 55 from a second academic-paper feed (counted separately from the academic papers), 31 company and personal blog posts, 9 subscription newsletters and 4 company filings. Of those 432, 115 were read in full and 145 were screened out as off-topic. Another 3 were logged with a stated reason for not reading closely, and 1 needs to be fetched again. The other 168 haven't been read yet; all are papers, 113 academic papers and 55 from the second paper feed. The 145 screened out are 47 academic papers, 76 social-platform posts, 14 blog posts, 6 newsletters and 2 company filings.