ISO-NE large loads: not yet at scale in New England — FERC EL26-72 show-cause, 90-day abeyance, and a 16 November Section 205 tariff clock.
What happened: ISO New England’s Large Loads Integration Key Project page frames data centers and other large loads as a reliability and interconnection design problem that has not yet materialized in New England at scale. Clock locked: on 18 June 2026, FERC issued a show-cause order under Section 206 of the Federal Power Act (EL26-72-000) on integration of large loads and co-located load arrangements. Process locked: on 14 August 2026, FERC granted a 90-day abeyance; ISO-NE will run a stakeholder process ahead of a 16 November 2026 Tariff filing under Section 205 covering the five reform categories identified in FERC’s order. Adequacy locked: the show-cause also required an informational report on resource adequacy; ISO-NE’s 20 July 2026 filing told FERC it is developing resource-adequacy-related provisions for large-load interconnections and adverse reliability risks. Scale language locked only as the page’s “not yet materialized … at scale” sentence — not a queue census and not a TWh path. This is a New England stakeholder process and sixth-RTO tariff clock — not MISO’s 120-day / ZGIA path, not SPP’s CHILL/HILLGA 90-day path, not CAISO’s EL26-71 clock (draft final 24 September / Board 28 October), not PJM ride-through after nearly 4,000 MW of disconnection, not ERCOT Batch Zero queue megawatts, and not IEA or LBNL national TWh shares.
Why it matters: After Midwest speed claims, Plains conditional service, West-coast tariff clocks, and ride-through design, the environment beat is how New England still says large loads have not arrived at scale — with a November Section 205 filing still ahead.
NY Fed Liberty Street: job postings show little distinct AI-driven hiring drop — relative decline in high-exposure vacancies predates ChatGPT.
What happened: Federal Reserve Bank of New York Liberty Street Economics post by Audoly, Guerin, and Topa, Do Job Postings Show Early Labor-Market Effects of AI? (14 May 2026; DOI 10.59576/lse.20260514), combines an Anthropic occupational AI-exposure metric (O*NET tasks × observed usage; automate vs augment) with U.S. Lightcast job postings in an event study around ChatGPT in late 2022. High-exposure cutoff locked at ≥0.2 (results similar under alternative cutoffs); occupation weights from 2019 vacancies. Exposure locked: less than 10 percent of workers and vacancies sit in occupations with AI exposure of at least 0.4; 40 percent of workers are in jobs with zero measured AI exposure (employment distribution 2024; vacancies January 2026). Result locked: overall hiring has slowed since ChatGPT, but postings provide little indication of a distinct AI-driven decline in labor demand. The relative decline in vacancies for higher-exposure occupations predates ChatGPT; divergence began before 2022 and does not show a clear additional break after 2022. The gap between high- and low-exposure jobs stabilizes after 2023. Junior vs senior postings within high-exposure occupations move broadly in parallel — the slowdown is not concentrated in entry-level highly exposed jobs. Conclusion locked: AI may be contributing but is not the main driver of the hiring slowdown. This is a staff blog on national Lightcast postings — authors’ views, not FOMC policy, not yesterday’s Crane–Soto FEDS 2026-018 coder deceleration paper, not Dallas Fed’s Texas Lightcast −5% / ~−8% within-industry automatable-task print, not NY Fed regional 61%/51% adoption packages, and not a mass-layoff census.
Why it matters: After coder occupation papers and state-level automatable-task prints, the jobs beat is a national postings event study that puts the relative drop in AI-exposed vacancies before ChatGPT and then flat — not a pink-slip count.
CISA and NIST finalize IR 8587: harden cloud identity tokens against theft and forgery — with high-level AI and post-quantum notes.
What happened: CISA’s news release CISA and NIST Release Guidelines to Protect Federal Cloud Identity Systems from Token Theft, Forgery, and Misuse and NIST’s companion news finalize IR 8587, Protecting Tokens and Assertions from Forgery, Theft, and Misuse: Implementation Recommendations for Agencies and Cloud Service Providers. Instrument locked: final report incorporates feedback from nearly 250 public comments on token validation, secrets management, and detection at scale; builds on NIST SP 800-53 Release 5.1.1 control IA-13; responds to Executive Order 14306. Incident locked from NIST news: the report cites an attack in which foreign actors accessed agency email using forged tokens from a single stolen commercial signing key and stole more than 60,000 emails from a single agency. AI lock: authors added new high-level considerations for handling AI and migration to post-quantum cryptography (PQC); the publication does not offer a comprehensive toolkit for either topic. Status locked: final interagency implementation report — not a Binding Operational Directive in this extract and not a statute. Audience locked as federal agencies and cloud service providers; any organization using tokens in access management can look to it. This is an identity-token / SSO hardening beat with an explicit AI + PQC add — not yesterday’s CISA Careful Adoption of Agentic AI Services guide, not NIST’s AI RMF Critical Infrastructure profile, not NIST AI 200-2, not SP 800-239 campus/data-center comments due 25 September, not the NVD modernization RFI due 13 October, and not a restatement of day-52 EU Article 50 duties (calendar only).
Why it matters: After agentic-adoption guides and critical-infrastructure profile notes, the policy beat is how agencies harden the tokens that open cloud identity systems — including a high-level AI and post-quantum paragraph, not a finished agent-control set.
Nature on Google Co-Scientist: autonomous agents combed more than 700 papers, generated 108 strategies, and kept one — glue condensates, don’t dissolve them.
What happened: A Nature technology feature, AI co-scientists are revolutionizing how research is done, recounts Google’s Co-Scientist system launching several autonomous agents to search and synthesize papers, evaluate competing explanations, refine and critique hypotheses, and iteratively test ideas against published evidence — not a single-pass chatbot. Case locked (Whitehead Institute / Pertl–Overholt / Young; MYC condensates): after about an hour of correction (the model first treated protein clusters as the drug rather than the target, and assumed clusters always activate expression), Co-Scientist combed more than 700 scientific papers and generated 108 possible strategies; it rejected all but one. Remaining strategy locked: glue condensates together with click chemistry rather than dissolving them (Dewpoint’s direction) so MYC DNA can no longer be read. Overholt is quoted that the laboratory “had certainly never thought about anything like this.” Prior lab examples locked as feature illustrations only: a drug combination that kills leukaemia cells in a dish, and a treatment that in the lab regenerates liver tissue damaged by disease. Limits locked: journalism feature — not a peer-reviewed trial metric and not a bedside authorization; the hypothesis still needs human verification. This is a hypothesis-generation workflow recap — not yesterday’s AISB/OpenFold3 proprietary protein–ligand lift (20,167 / 1,056), not Virtual Biotech’s multi-thousand-agent discovery swarm, and not Paper2Agent paper-to-agent conversion rates.
Why it matters: After closed structural-biology model lifts and agent-swarm discovery headlines, the science beat is how multi-agent co-scientist systems propose testable strategies humans had not named — still a feature, not a clinical endpoint.
UNESCO IESALC: across 200 LAC higher-ed institutions in 19 countries, 87% use AI — only 26% have a formal strategy.
What happened: UNESCO IESALC article New UNESCO IESALC study reveals widespread AI adoption in higher education across Latin America and the Caribbean, while governance lags behind (launched during Digital Learning Week 2026) reports the study AI Implementation in Higher Education in Latin America and the Caribbean, produced with UNU-IAS. Census locked: data from 200 higher education institutions across 19 countries. Uptake locked: 87% of institutions use AI in at least one area; 74% in teaching and learning (often general-purpose tools such as ChatGPT, Copilot, and Gemini); 57% in research (data analysis / research quality). Governance locked: only 26% have a formal AI strategy; even fewer have clear policies, governance structures, and monitoring/evaluation. Adoption is often driven by faculty, researchers, and students rather than institutional policy. Staff training is common; student training lags — 1 in 2 students reporting confusion about appropriate AI use. Equity locked: private institutions tend to lead on strategy, staff training, and governance; public institutions face greater resource and capacity constraints. Limits locked: institutional self-report census — not an RCT, not a K–12 enrollment print, and not PISA. This is a higher-ed adoption-versus-governance census — not UNESCO’s algorithm in the room think-piece volume, not the LAC Observatory learning-crisis platform, not World Bank Mongolia TVA lesson-plan counts, and not UNICEF’s child-uptake snapshot.
Why it matters: After governance essays and national teacher-tool rollouts, the education beat is a regional university census: AI is already in classrooms and labs, while formal strategy and student guidance lag.
A New England not-yet-at-scale clock is not a Midwest 120-day path.
What happened: Keep ISO-NE as FERC 18 June 2026 / EL26-72-000 show-cause framing; 14 August 90-day abeyance; 16 November 2026 Section 205 tariff filing across five reform categories; 20 July resource-adequacy report; large loads not yet materialized in New England at scale. No queue MW census and no TWh locked. Keep separate from MISO 120-day / ZGIA, SPP CHILL/HILLGA 90-day, CAISO EL26-71 / 24 September draft final, PJM ride-through ~4,000 MW, ERCOT Batch Zero, IEA +17% 2025, IEA Grids queue totals, and LBNL share.
Why it matters: Regional interconnection design is a different evidence class than metered growth rates or single-event disconnection reports.
A pre-ChatGPT relative posting drop is not a coder pink-slip count.
What happened: Keep NY Fed LSE 14 May 2026 as national Lightcast + Anthropic exposure event study; <10% of workers/vacancies at exposure ≥0.4; 40% zero exposure; high-exposure cutoff ≥0.2; relative decline predates ChatGPT; gap stabilizes after 2023; no junior/senior split; AI not the main driver of the hiring slowdown. Keep separate from FEDS 2026-018 coder deceleration, Dallas Fed Texas −5%/−8%, NY Fed 61%/51%, Philadelphia Fed LIFE, Minneapolis Fed Freund–Mann, Boston Fed CPP 26-8, and St. Louis RPS shares.
Why it matters: A national postings event study is a different evidence class than occupation-shock working papers or household fear surveys.
Token/SSO hardening is not an agentic-adoption control set.
What happened: Keep NIST IR 8587 as final token/assertion hardening with CISA; nearly 250 comments; SP 800-53 IA-13; EO 14306; Storm-0558 more than 60,000 emails from one agency; high-level AI + PQC considerations, not a comprehensive AI toolkit. Day-52 EU Article 50 remains calendar only. Standing clocks only: Canada ISED closes 23 September; FTC personalized pricing and NIST SP 800-239 still due 25 September; NIST AI 200-2 still due 6 October; NVD modernization RFI still due 13 October; UNESCO consultation still due 15 October; FDA GenAI-device comments still due 19 October; CAISO Board still 28 October; ISO-NE Section 205 still 16 November.
A co-scientist hypothesis workflow is not a clinical endpoint trial.
What happened: Keep Google Co-Scientist as multi-agent paper search/synthesis and critique; more than 700 papers; 108 strategies; one remaining (glue condensates via click chemistry vs dissolve); MYC/cancer case; leukaemia-in-a-dish and lab liver-regen as prior feature examples only. Keep separate from AISB/OpenFold3 20,167 / 1,056 and from Virtual Biotech multi-thousand-agent discovery.
Why it matters: Hypothesis-generation features are a different evidence class than proprietary structural lifts or bedside RCTs.
Eighty-seven percent uptake with 26 percent strategy is not a learning-outcomes print.
What happened: Keep IESALC as 200 institutions / 19 countries; 87% any-area AI; 74% teaching/learning; 57% research; 26% formal strategy; 1 in 2 students confused on appropriate use; private lead public on governance. Standing context only: UNESCO consultation comments due 15 October. Keep separate from UNESCO algorithm-in-the-room think pieces, LAC Observatory, Mongolia TVA lesson-plan counts, and UNICEF child-uptake snapshots.
Why it matters: An institutional self-report census is a different evidence class than governance essays or child-protection uptake prints.
The full daily ledger keeps broader source-linked coverage organized by topic. Story dates are shown separately from the September 23 edition date.
September 23 · ISO-NE large loads
ISO-NE Large Loads Integration Key Project: FERC 18 June 2026 show-cause EL26-72-000 (Section 206); 14 August 90-day abeyance; Section 205 Tariff filing 16 November 2026 across five reform categories; 20 July resource-adequacy report; large loads not yet materialized in New England at scale — not a queue census.
New England stakeholder process + tariff clock — not MISO 120-day / ZGIA; not SPP 90-day CHILL/HILLGA; not CAISO EL26-71; not PJM ride-through.
Show-cause and abeyance frame co-located load arrangements and resource-adequacy risks; ISO is developing provisions to facilitate interconnections while protecting reliability — process design, not delivered megawatts.
Conditional interconnection design — not ERCOT Batch Zero; not LBNL TWh share; not IEA +17% 2025.
NY Fed Liberty Street (Audoly–Guerin–Topa, 14 May 2026): Lightcast postings × Anthropic exposure; <10% of workers/vacancies at exposure ≥0.4; 40% zero exposure; relative high-exposure vacancy decline predates ChatGPT and stabilizes after 2023; AI not the main driver of the hiring slowdown.
Staff blog — not FEDS 2026-018; not Dallas Fed Texas −5%/−8%; not NY Fed 61%/51%; not FOMC.
Within high-exposure occupations, junior and senior postings move broadly in parallel — the slowdown is not concentrated in entry-level highly exposed jobs. High-exposure cutoff ≥0.2; employment distribution 2024; vacancies January 2026.
Postings event study — not Philadelphia Fed LIFE; not Minneapolis Fed Freund–Mann.
NIST IR 8587 final with CISA: Protecting Tokens and Assertions from Forgery, Theft, and Misuse; nearly 250 public comments; SP 800-53 IA-13; EO 14306; cites attack stealing more than 60,000 emails from one agency via forged tokens; high-level AI + PQC considerations.
Identity-token / SSO hardening — not CISA agentic-AI adoption guide; not NIST CI RMF profile; not SP 800-239; not NVD RFI.
NIST news on IR 8587 finalization: implementation recommendations for agencies and cloud service providers; AI and post-quantum paragraphs are high-level only, not a comprehensive toolkit. Final interagency report — not a BOD.
Token hardening — not AI 200-2; not NCCoE agent-identity restamp.
Nature technology feature on Google Co-Scientist: multi-agent paper search/synthesis and critique; more than 700 papers; 108 strategies; one remaining (glue MYC condensates via click chemistry rather than dissolve); leukaemia-in-a-dish and lab liver-regen as prior examples; not a peer-reviewed trial.
Hypothesis-generation workflow — not AISB/OpenFold3 20,167 / 1,056; not Virtual Biotech ~37k agents; not a bedside RCT.
UNESCO IESALC (with UNU-IAS): 200 higher education institutions across 19 LAC countries; 87% using AI in at least one area; 74% teaching/learning; 57% research; only 26% formal AI strategy; 1 in 2 students confused about appropriate use; private institutions lead public on governance.
HE adoption-vs-governance census — not algorithm-in-the-room 26/six; not LAC Observatory; not Mongolia TVA; not UNICEF ≥20 million / >3×; not PISA.
Day-52 EU Article 50 calendar only; Canada ISED closes 23 September; CAISO draft final still 24 September; FTC personalized pricing and NIST SP 800-239 still due 25 September; NIST AI 200-2 still due 6 October; NVD RFI still due 13 October; UNESCO consultation and SPP Q&A still due 15 October; FDA GenAI-device comments still due 19 October; CAISO Board 28 October; ISO-NE Section 205 16 November.
