Leadership & Executive
The Board Room
Microsoft fixed an exploited 10.0 in Entra ID and told customers no action was required.
Remote code execution against the identity control plane is the one failure mode where tenant-side controls contribute nothing. There was nothing to patch, and no way to confirm whether a given tenant was touched in the window before disclosure. Which means the answer you hand an auditor about that period is whatever the vendor decides to publish, not anything you can produce yourself.
In Play
Four Criticals, Two With No Patch You Could Apply
The Hacker News reported that Microsoft disclosed a maximum-severity remote code execution flaw in Entra ID (CVE-2026-69836), already under exploitation, and told customers no action was required. Your identity control plane had nothing to patch and no tenant-level way to verify exposure. Three more criticals landed in the same window: five Cisco 10.0s, a GitLab flaw weaponized within days, and Rust crates with 244M downloads carrying build-time malware.
Ask ClarityPer-Task AI Costs Are Rising While Per-Token Prices Hold
Grok 4.6 scored 1,577 Elo against Claude Opus 5's 1,715 on Artificial Analysis' AA-Briefcase test of multi-week knowledge work — and got there in roughly half the turns and a quarter of the input tokens. For any long-horizon agent in production, the cheaper model is now the one that finishes in fewer round trips. Per-task costs are climbing even as list prices hold: Grok 4.6 gained five index points and moved from $0.36 to $0.84 per task.
Ask ClarityTexas Halts Data Centers as Opposition Goes Cross-Partisan
Governor Abbott says his directive halted up to 1,800 data center projects in Texas, and local opposition now runs near 75% with no meaningful partisan split. Moratorium campaigns are live in Michigan and Pennsylvania. If your 2027-2029 capacity plan still assumes 2024 permitting speed, schedule is now the binding constraint rather than chips. Bloomberg reports that Fluidstack, which builds Anthropic's data centers, now hires staff to face organized opposition before capital is committed.
Ask ClarityThe Biggest AI Buyers Are Each Other
Meta has become one of Microsoft's largest AI customers, and Bloomberg reports AI demand remains concentrated inside the tech industry itself. Alibaba published what that dynamic costs: profit fell more than 75% while quarterly capital spending reached almost $10 billion — spending it framed as defending position rather than generating return. If your AI pipeline traces back to other tech companies' capex budgets, that is one correlated exposure wearing the costume of a diversified book.
Ask ClarityWhere Your Work Came From Becomes Checkable
Anthropic is embedding SynthID-Text watermarks in every Claude model launched after August 2, 2026 — worldwide, not only in the EU — to satisfy Article 50 of the AI Act, with a detection API to follow. Within 12-18 months a customer, auditor, or acquirer can estimate the AI-generated share of your codebase or your consulting deliverables. Separately, Simile AI's $2B Series B puts a price on the first-party behavioral data most companies signed away in older analytics contracts.
Ask Clarity
Deep Dives
- ●
Two of the Four Criticals Had Nothing You Could Patch
The constraint is not awareness — it is how much remediation your platform team can push through, and what your identity vendor is contractually obliged to prove.
One manifest line was the entire attack The largest Rust crate compromise by download count needed no exploit. One dependency line went into the Cargo manifests of arrayref (244M downloads) and append-only-vec (4M), pointing at proc-macro1 , a typosquat that…
3 action items
- ●
Per-Token Prices Are Flat. Per-Task Costs Just Doubled.
The model that wins your workload is the one that finishes in the fewest round trips, and the vendors posting the biggest capability gains are billing you for the extra ones.
Capability is being funded by output length The two largest capability gains this cycle were bought with tokens, not efficiency. Qwen3.8-Max added eleven points on Artificial Analysis' intelligence index and moved from $0.54 to $1.13 per task, entirely because it…
3 action items
- ●
Your Patent Says Five Humans. Your Press Release Said the AI.
Three moves gave outsiders a way to check where your work came from — and put a price on the one input none of them can synthesize.
The seam opposing counsel will find first Insilico Medicine said a pulmonary fibrosis molecule was "discovered by" generative AI , then filed the patent naming five humans and no AI . Only humans can be inventors, whatever the model contributed,…
3 action items
The edition continues
Take the signal into the room.
Sign up or log in to read all 3 deep dives in full, plus the final take.
Read the full editionContinue with LinkedIn