Spotlight features, sponsorships, reports, advisory — tell us who you are and what you are thinking. No plans, no pricing sheets.
Weekly insights on working, building, and making AI part of how you operate — from my own builds and from others doing the same.
Thirty stories from the week capacity, money and permission all pushed back: the vote, the selloff, the IPO and the security gate.
The week began with a vote and ended with a repricing. Voters restricted data centers in 40 of 41 local measures. The Nasdaq 100 fell 6.3% in a day. Anthropic’s IPO became the first public test of AI revenue.
Thirty stories explain why the constraint is no longer the model. They are sorted into five sections: capacity and permission, markets and monetization, the security frontier, agents in production, and models, products and distribution.
The short version
Ballotpedia tracked 50 local data-center measures in 13 states in 2026; 41 were on the Nov 3 ballot. Voters favored limiting development in 40 of 41 contests, and in 46 of 47 decided measures across the year. The one Nov 3 exception was Harwood, North Dakota, where voters rejected an additional review requirement.
Most builders will never site a data center. They will still feel this in quotas, lead times and price. Local permission is now a supply-chain input.
Why it mattersA moratorium can outlast a model generation. Plan for permission the way you plan for compute.
CNBC counted 41 data-center measures in 10 states. Voters backed restrictions or rejected pro-development measures in all but one.
Why it mattersThe business press read the vote as a market signal, not a local fight.
cnbc.com →Voters approved nearly every data-center moratorium on the ballot, dozens in total.
Why it mattersMoratoria stop the clock, which is the one resource the buildout cannot buy.
theguardian.com →The trade press confirmed a wave of local moratoria, limits and restrictions in the Nov 3 votes.
Why it mattersWhen the industry’s own press leads with permission risk, operators can no longer treat it as activism noise.
datacenterdynamics.com →Hundreds of candidates who pledged to rein in data centers won office.
Why it mattersThe constraint now has a four-year horizon; a moratorium can outlast a model generation.
nbcnews.com →The rejection was not concentrated in one state or one party.
Why it mattersThat makes it a planning assumption, not a regional risk.
stateline.org →The Nasdaq 100 fell 6.3% on Nov 5, erasing $2.3 trillion. Nvidia fell 9.2%, and the Philadelphia Semiconductor Index fell 12.3%.
Why it mattersThe selloff was triggered by weak chip earnings and amplified by the midterm result.
nbcnews.com →Five companies, including SpaceX, are set to spend up to $828 billion on capex in 2026. Four of the five had negative free cash flow in Q3.
Why it mattersAnthropic alone disclosed $265 billion in future minimum commitments.
fortune.com →The five hyperscalers had already spent $556.2 billion in the first three quarters of 2026.
Why it mattersThe money is going out the door before the revenue is proven.
businessinsider.com →Buyers valued Anthropic as low as $1.1 trillion, down from a $1.5–2 trillion range discussed earlier.
Why it mattersThe IPO is the first public mark on a frontier lab.
bloomberg.com →Three major investors said Anthropic will likely need to cut its $1.5 trillion target, with one suggesting about $1 trillion.
Why it mattersThe skepticism is about the multiple, not the product.
wsj.com →The amended S-1 shows $16.96 billion in Q3 revenue, up 47% sequentially, and a $567 million operating loss. Purchase and lease commitments are about $895.4 billion, mostly compute.
Why it mattersValuation talks are $1.1–1.3 trillion, down from $1.5–2 trillion. The bill is 13 times the run rate.
finance.yahoo.com →Nvidia opened an investigation into its CEO’s relationship with a person alleged to be a Chinese spy.
Why it mattersThe most important supplier in the buildout is now a governance risk.
theverge.com →GPT-6 Cyber helped develop 948 Linux privilege-escalation exploits and find 81 previously unknown high-severity Firefox vulnerabilities. It avoided out-of-scope services in 99.93% of tests without a deployment prompt and 100% with it.
Why it mattersOpenAI classifies it as Critical. The gate is part of the product.
openai.com →Reuters confirmed the model went to a small group of vetted customers, not the general public.
Why it mattersThe gate is part of the product.
reuters.com →GPT-6 Cyber scored 48.5% on ExploitGym, about 16 points above GPT-5.6 Cyber. The Critical label means it can develop zero-day exploits in hardened systems without human intervention.
Why it mattersEvaluation awareness remains a concern.
theverge.com →Independent security coverage confirmed the 948 Linux exploit and 81 Firefox vulnerability claims.
Why it mattersThe numbers held across outlets.
thehackernews.com →The Register framed the release around scope adherence, not raw capability.
Why it mattersFor practitioners, staying in bounds matters more than a benchmark score.
theregister.com →GPT-6 Astra was the first OpenAI model to cross the Critical threshold in September. November made that capability an access-controlled product.
Why it mattersThe trend is the gate, not the one model.
csoonline.com →Only 9.5% of organizations have a named person formally accountable for agent behavior. 52% had an agent security or privacy incident in the past year. 6% could revoke a rogue agent’s access within an hour.
Why it mattersThe market is repricing compute; most organizations still cannot name who answers for an agent.
gravitee.io →Gravitee’s analysis: agents are in production, accountability is not. Self-assessed readiness fell even as capability rose.
Why it mattersA clean secondary write-up if the main report is too dense.
gravitee.io →Organizations running more than 150 agents more than doubled since April, to 16% from 7%.
Why it mattersThe estate is scaling faster than the control layer.
businessinsider.com →Fiddler puts agent failure rates in production at 70% to 95%, driven by compounding errors and tool breakdowns.
Why it mattersUse it as a range with attribution, not a universal law.
fiddler.ai →The Institute PM reports 88% of AI agent pilots never reach production. The blockers are integration, reliability, latency and security, not model quality.
Why it mattersA direct challenge to the “better model solves it” assumption.
institutepm.com →Digital Applied separates adoption claims from production definitions and ROI estimates.
Why it mattersAdoption numbers depend on what counts as production.
digitalapplied.com →Gemini 4 Argon launched at $2 per million input tokens and $10 per million output tokens through Nov 30. Standard pricing from Dec 1 is $4 and $20.
Why it mattersThe price of frontier intelligence is scheduled to double in three weeks.
cloud.google.com →DemandSphere’s tracker compares benchmarks, pricing and capabilities across frontier models.
Why it mattersThe shelf changes monthly, which is the story.
demandsphere.com →Four frontier models shipped on Sept 22, and seven Claude models were retired during 2026.
Why it mattersThe real cost is migration and portability, not benchmark rank.
ibl.ai →OpenAI says ChatGPT handles roughly one-fifth of Google’s query volume, with 1.2 billion weekly active users. Outside estimates are 7.5% to 12%.
Why it mattersPresent the claim and the gap, not just the headline.
forbes.com →DeepSeek’s open-source agent app, Harness, reached 1.54 million daily active users by Nov 4, less than three months after its Aug 13 launch.
Why it mattersCheap agent tooling is scaling while the frontier gets gated.
bloomberg.com →The operator test
What breaks first at 10x, and what will you do about it? The feature turns the week into a three-part test for your own roadmap.
Read the feature: The Morning After the Hype
Watch the full session: Morning Scrum with Govind Davis, Nov 10 2026
New articles, show notes, and practical lessons from the Strattegys AI crew.
InsightsA Morning Scrum field note on using AI to produce a Strattegys Spotlight package, and building the missing production tools as the work unfolded.
FeaturedA vanilla Node server, four agent consumers, and the rules that keep capabilities safe: least privilege, one builder, and receipts for everything.
FeaturedThe AI superpower arc hits the scale wall. A feature on drawing the line — Ghost Studio, ScreenPAL, and the rebuild question.
Get your founding sponsor spot now.
Inquire to learn about features, shows, and sponsor placements. No plans, no pricing sheets — just a conversation about fit.
