Durable transitions connecting signals, capability deltas, projects, systems, and counter-evidence.
/shifts.jsonOpen JSON indexThis week
The weekly knowledge desk published 9 reviewed updates on Oct 26, 2026.
OpenAI tightens Pro quotas as serving costs vary by provider
Glean makes Agent Identity generally available for governed agent accounts
Cloudflare’s typed extraction model shifts work from JSON retries to schema policy
Durable agent jobs and shared Team Tasks put work into persistent workflows
Last week’s items point to typed and bounded AI decision interfaces
Durable agent harnesses and bounded loops surface in last week’s coverage
IBM describes learned behavioral guidance that can improve an agent without changing model weights
LangChain guide and middleware paper point to memory beyond individual AI clients
New Runtime items point to review and validation as coding agents produce more
Auto-created by the weekly desk
Drafted from reviewed evidence and published without a site rebuild. They join the catalog below at the next release.
Manual OCR pipelines with hand‑written regex/JSON schemas and static error handling -> LLM‑guided OCR and schema creation via conversational interfaces, with dynamic error correction
9 shifts
AI systems are gradually moving beyond text answers and assembling verifiable interfaces around a user's specific intent.
Autonomous agents are moving from borrowed user sessions and shared keys toward distinct identities with scoped authority, sponsors, expiry, and audit.
Context is shifting from private memory inside one AI client into a governed layer between agents, IDEs, documents, and tools.
As agents produce more implementation, engineering responsibility is moving toward specifications, evidence, acceptance decisions, and release accountability.
Coding agents are expanding beyond software implementation into context-aware workbenches that assemble prototypes, interfaces, documents, workflows, and operational artifacts.
Knowledge work is beginning to route through durable goal objects that assemble temporary human-agent-tool configurations while preserving authority, evidence, and accepted state.
Long-running agent work is moving from manual continue prompts toward explicit goals, limits, subagents, schedules, and verifiable completion.
The AI API is becoming not just a model interface, but a product contract around scarce compute, limits, and degradation.
Organizations are moving the durable learning asset out of one model provider and into user-owned traces, evaluations, corrections, and promotion rules.
No shifts match these filters.
How a capability moves from spectacle to infrastructure.
GET /shifts.jsonGET /shifts/[slug].mdGET /shifts/[slug].jsonLoading the privacy-safe route aggregate…
Open the JSON contract