Format
Models
Step-5 Preview lists $0.05 per million for re-reading context – 4x below the $0.20 tier Anthropic, Meta, and Google occupy on the same axis. The missing write terms turn the arithmetic into a risk calculation.
◆ Lena Park
Agents at Work
The third major IT services firm in 48 hours announces a named production customer for agentic AI. Only one of the three has shared performance data.
◆ Dana Ellison
Trust & Security
The management interface of IBM Security Verify Access processes untrusted data in a way that allows an unauthenticated remote attacker to execute arbitrary code. It is not an authentication bypass; the system processes the malicious payload directly.
◆ Heath Callahan
Commerce
The first major L1 to embed programmable wallets for AI agents bypasses the merchant-of-record debate by building underneath it. The accountability gap, however, remains wide open.
◆ Tessa Vaughn
Commerce
Amex's "Proof of Intent" token is the sixth definition of "agentic checkout" and the first from the issuer side. It moves the evidence issuers hold, not the merchant-of-record loss default.
◆ Tessa Vaughn
Agents at Work
In roughly 48 hours, HCLTech and DXC have both pivoted from selling IT labor to selling the guardrails around AI agents. DXC's release prints the wave's first outcome figures: 95% faster resolution across 500 use cases, self-measured, no customers named.
◆ Dana Ellison
Agents at Work
The Verge reports Shopify is opening more than a billion listings to shopper-configured agents with budget rules and per-purchase approval. Shopify's own newsroom shows the same direction: any AI channel, any brand, merchant of record unchanged.
◆ Dana Ellison
Infrastructure
The draft gives builders a protocol to measure against. What is missing tells them where the real risk sits.
◆ Blair Hayes
Models
Google's flagship model goes from gated cyber-defender access to general availability at $2/$10 per million tokens — undercutting the $4/$20 frontier floor that Anthropic and Meta just established. Standard pricing arrives December 1.
◆ Lena Park
Trust & Security
Defenders who relied on Ivanti's September batch bulletin remain vulnerable to an actively exploited authentication bypass (CVE-2026-104935) that requires an emergency firmware upgrade. Federal agencies have 72 hours.
◆ Heath Callahan
