Tuesday, September 22, 2026 · Daily edition
MISO’s 120-day large-load path, a Fed paper on coder slowdown, and CISA’s agentic AI guide
Today’s ledger follows the Midwest grid operator’s fast path for data-center and AI loads, a Federal Reserve staff working paper that finds coder employment slowed sharply after ChatGPT, joint cybersecurity guidance on careful agentic AI adoption, a Nature recap of drug firms pooling private protein–ligand data, and UNESCO’s questions about agency when algorithms enter the classroom.
MISO’s large-load path: approval within 120 days, co-located generation, and reliability tradeoffs
What happened. Midcontinent Independent System Operator’s Large Load Additions page frames an unprecedented surge in interconnection requests from data centers, advanced manufacturing, AI computing facilities, and other energy-intensive industries. Many customers want to connect within 18 to 36 months — faster than traditional generation and transmission build timelines. On 18 June 2026, FERC directed all RTOs and ISOs, including MISO, to update large-load tariff provisions. MISO’s answer is a three-part framework: connect quickly, with large loads able to get approval within 120 days once studies and generator interconnection agreements are signed; transition with flexibility through co-located resources, non-firm service, and firm step-ups; and operate for long-term reliability with telemetry, curtailment, ride-through, and stability controls. Named tools include ZGIA co-location, expedited generation (ERAS), and fast load studies (EPR). The page speaks only of potential to add thousands of megawatts — not a queue census and not a national electricity path.
What to watch. After Plains conditional service and West Coast tariff clocks, the Midwest story is speed sold through a 120-day approval claim and co-located generation tools. That is a reliability bargain, not another metered growth print — and it is not SPP’s 90-day conditional path.
Read the MISO Large Load Additions page →
Fed staff paper: coder employment decelerated sharply since ChatGPT — still growing, not like before
What happened. Board of Governors staff economists Crane and Soto ask whether large language models have left a visible mark on employment yet, focusing on computer programming–intensive occupations because coding is among the most LLM-exposed tasks. In FEDS 2026-018, AI and Coder Employment: Compiling the Evidence, they link O*NET to CPS and find that aggregate employment of coders has decelerated sharply since ChatGPT arrived. A novel industry-level shock control shows the slowdown is not just coders sitting in slowing industries — it is an occupation-specific shock around the model release. Coder employment has continued to grow in recent years, but much more slowly than before 2022. The authors say the results suggest a measurable and potentially consequential impact for some groups, while noting they cannot settle general-equilibrium demand effects. It is a staff working paper on one occupational group — authors’ views, not FOMC policy and not a mass-layoff census.
What to watch. After buildout dashboards and whole-market perception splits, the jobs beat is a Board staff paper that isolates coder slowdown as occupation-specific. Still growing — just not on the pre-ChatGPT path.
Read the Board FEDS paper →
CISA and partners: careful adoption of agentic AI — start low-risk, avoid unrestricted access
What happened. CISA’s resource page Careful Adoption of Agentic AI Services (1 May 2026) and a same-day news release with Australia’s ACSC and other U.S. and international partners warn that critical infrastructure and defense are increasingly deploying agentic AI. The pages name extra risks: expanded attack surface, privilege creep, behavioral misalignment, and obscure event records. The practical instruction is blunt — avoid granting broad or unrestricted access, especially to sensitive data or critical systems; begin with low-risk, non-sensitive use cases; and fold agentic AI security into the organization’s existing risk model. Status is joint guidance, not a Binding Operational Directive and not a statute.
What to watch. After critical-infrastructure profile concept notes, the policy beat is how operational cybersecurity agencies tell operators to adopt agents without handing them the keys on day one.
Read the CISA agentic AI guide →
Nature on AISB / OpenFold3: private protein–ligand data lifts held-out accuracy — still unpublished and closed
What happened. A Nature news recap describes the AI Structural Biology Network — AbbVie, Astex, and several other drug companies — fine-tuning OpenFold3 on a further 20,167 structures of proteins bound to potential drugs or ligands from five companies, with proprietary data kept private. On 1,056 held-out protein–ligand structures, the AISB model predicted more than half to a high level of accuracy; publicly available OpenFold3 hit that bar on about one-third; Boltz-2 about 40%. The pooled model also beat tools trained only on each company’s individual data. Limits are plain: described in a blog post, not peer-reviewed, and the model is not publicly available. The team plans to submit a paper. Public-data efforts such as UK OpenBind, supported by up to £8 million, are named as a parallel track.
What to watch. After ethics-oversight guidance and discovery headlines, the science beat is how pooled proprietary binding data can lift protein–ligand models — with the honesty that the lift is still unpublished and closed.
Read the Nature recap →
UNESCO’s “algorithm in the room”: 26 think pieces and six questions on agency and education as a common good
What happened. UNESCO’s article The algorithm in the room (8–9 September 2026) introduces a compilation of 26 think pieces on education in an era of AI and algorithms. Six urgent questions run through the set: who will hold agency, autonomy, and oversight; how young people will develop relationally, emotionally, and cognitively; what will remain meaningful for pedagogy and assessment; how we will think, question, and express ourselves; how we will govern education as a common good accountable to the public; and how AI might reshape imagination about education’s future. It is a think-piece volume — not a randomized trial, not an enrollment census, and not a student-uptake print.
What to watch. After national teacher-tool architecture and child-protection snapshots, the education beat is the harder public questions — who keeps agency when algorithms enter the room.
Read the UNESCO essay →
Also in today’s ledger
• A 120-day Midwest path is not a Plains 90-day clock — and not a terawatt-hour census.
• Coder slowdown is not a whole-market pink-slip count — and joint agentic guidance is not a finished control set.
• A proprietary-data lift is not a clinical endpoint trial — and twenty-six think pieces are not a learning-outcomes census.
|