Daily Positive Signals: a cell-by-cell DNA reader, climate models in seconds, and a 1,000-qubit quantum plan

On this page
  1. A DNA model that reads variants cell by cell, not tissue by tissue
  2. Climate emulators that learn from one scenario instead of six
  3. IBM’s road to a 1,040-qubit quantum system
  4. New York turns AI safety promises into law
  5. Useful progress, without the hype
  6. Know someone this would help?
  7. References

Listen to this brief · about 6 min, story text only

This edition: DeepMind’s AlphaGenome 2 reads DNA variants cell by cell; MIT builds trustworthy climate emulators from a single optimized scenario; IBM charts a path past 1,000 qubits; and New York turns AI safety promises into law. The thread: verification is becoming the product, from the lab bench to the statute book.

Signal 1 of 4 · Biotech

A DNA model that reads variants cell by cell, not tissue by tissue

107M+ cell profiles
individual cell profiles across 68 tissues used to train the model
30% vs 16%
causal variants found on the TenK10K immune benchmark, nearly double the earlier model

Google DeepMind announced AlphaGenome 2 on November 11, a DNA sequence model that predicts how genetic variants change gene activity in individual human cell types and states, rather than in tissue averages (Google DeepMind, 2026). The model takes a DNA sequence of up to a million letters plus a chosen cell type or state, and predicts gene expression, chromatin accessibility, histone marks, and DNA methylation for that condition (Google DeepMind, 2026). Trained on more than 107 million individual cell profiles across 68 tissues, it identified nearly twice as many causal variants as the original AlphaGenome on a benchmark of 28 immune cell types — 30% against 16%, with the direction of the gene effect right 43% of the time against 23% (Google DeepMind, 2026). It reproduced a known inflammatory bowel disease finding computationally, pinpointing the variant that overactivates the ETS2 gene in gut macrophages, and it is available now through the AlphaGenome API for non-commercial research (Google DeepMind, 2026; MLQ, 2026).

Why it matters. Most disease-linked variants sit outside protein-coding genes, and whether one matters depends on the cell it acts in. Pinpointing the variant, the gene, and the cell turns a statistical hint into a mechanism researchers can test.

Evidence status: DeepMind announcement and technical paper (November 11, 2026), independently reported by MLQ and Unite.AI. Limits: all performance figures are company-reported; the model is for research use only and has not been validated for clinical or diagnostic use; even at high precision it recovers only a minority of nominated causal variants.


Signal 2 of 4 · Climate

Climate emulators that learn from one scenario instead of six

1 scenario
optimized training scenario beats six standard emissions pathways
Seconds
for an emulator to run a climate scenario, against months for a full earth system model

MIT researchers have found a faster way to build trustworthy AI climate models, publishing the method in the Proceedings of the National Academy of Sciences on November 9 (Womack et al., 2026). Climate emulators are fast AI stand-ins for full earth system models, which can take months on a supercomputer to project a single future, while an emulator runs in seconds (Zewe, 2026). The team designed an algorithm that generates optimized, synthetic training scenarios, and an emulator trained on just one of them outperformed emulators trained on six standard emissions pathways (Womack et al., 2026). The optimized training also let the emulator separate the effects of greenhouse gases from aerosols, and capture the impact of local pollution controls that traditional emulators failed to predict (Zewe, 2026).

Why it matters. Building an emulator usually needs six to eight expensive full-model runs, so production takes months or years and the tools go stale. Needing just one run means updated emulators can reach the planners and researchers who rely on them far sooner.

Evidence status: Peer-reviewed study in PNAS (November 9, 2026); MIT News summary. Limits: demonstrated on a simple climate model and an intermediate-complexity model; testing on full-scale earth system models is the next step.


Signal 3 of 4 · Quantum

IBM’s road to a 1,040-qubit quantum system

1,040 qubits
planned 2027 system linking two Nighthawk r3 processors
19×
top speedups reported by partners running 34 billion circuits on Nighthawk r2

IBM announced its next quantum processor, Nighthawk r3, on November 11, designed to run 10,000 quantum operations across 520 programmable qubits, with 2,034 physical qubits on the chip (IBM, 2026). In 2027, IBM plans to link two r3 processors into a single 1,040-qubit system, using ten chip-to-chip connectors called l-couplers (IBM, 2026). The building blocks are already demonstrated: IBM linked Nighthawk r1 processors with four l-couplers at 99.3% state-transfer fidelity, toward a 99.9% target, and connected two cryogenic modules into one system for the first time (IBM, 2026). The current Nighthawk r2, in service since September, has run 34 billion circuits for partners including E.ON and Q-CTRL, with reported speedups of up to 19 times over earlier IBM processors (IBM, 2026; Ars Technica, 2026).

Why it matters. Single chips hit physical limits, so the path to useful quantum computers runs through linking chips together. IBM has now shown the links work, which makes the 1,040-qubit plan a build schedule rather than a wish.

Evidence status: IBM announcement (November 11, 2026), independently reported by Ars Technica and SiliconANGLE. Limits: r3 is a design target, not yet built; the 1,040-qubit system is a plan for 2027; partner speedups are company-reported.


Signal 4 of 4 · AI safety

New York turns AI safety promises into law

72 hours
to report a critical AI safety incident to the state
$1M / $3M
penalties for first and subsequent violations

New York opened a registration portal on November 10 under the RAISE Act, its frontier AI safety law, turning voluntary safety promises into legal duties (Bellan, 2026). Large frontier developers — companies with more than $500 million in annual revenue — must register by January 1, 2027, publish safety frameworks, issue public transparency reports, and report critical safety incidents within 72 hours (Bellan, 2026). Penalties reach $1 million for a first violation and $3 million for subsequent ones, enforced by the state’s new Office of Digital Innovation, Governance, Integrity, and Trust (Bellan, 2026). Governor Kathy Hochul also launched the New York Center for AI Responsibility and Research at Binghamton University, with $55 million in donations and capital, to independently evaluate advanced AI systems (Bellan, 2026).

Why it matters. Safety frameworks were voluntary pledges a lab could quietly rewrite. With California’s law already in force and Illinois’ taking effect in January, New York’s registration deadline makes a state-level floor for frontier AI accountability real. The November 1 deep dive asked whether a published rulebook can make AI safer — New York is now testing that question in law.

Evidence status: TechCrunch report (November 10, 2026), corroborated by multiple independent outlets. Limits: registration is the preparatory phase; full compliance duties begin in January 2027; no penalties have been reported; the content of the safety frameworks is still written by the companies themselves.

Editor’s note

Today’s thread is verification becoming the product. A DNA model proves its worth by finding the variant in the right cell, not just the right gene. A climate emulator earns trust by handling scenarios it never trained on. A quantum roadmap is credible because the links it depends on already work. And a safety law makes a lab’s promises checkable by someone other than the lab. That pairing — new power, plus a way to check it — is what makes progress durable rather than just fast.

That’s the edition — four signals, every claim verified against its sources. Browse past briefs · tomorrow’s lands at 07:15.

The Progress Brief, by email

Useful progress, without the hype

The daily Positive Signals brief and the weekly deep dive, by email. Evidence-checked, concise, no doom.

Free. One daily email, one weekly email. Unsubscribe anytime.

References

  1. Ars Technica. (2026, November 11). IBM to offer access to a dual-processor quantum system in 2027. https://arstechnica.com/science/2026/11/ibm-to-offer-access-to-a-dual-processor-quantum-system-in-2027/
  2. Bellan, R. (2026, November 10). OpenAI, Anthropic face new oversight as New York’s AI safety law takes effect. TechCrunch. https://techcrunch.com/2026/11/10/openai-anthropic-face-new-oversight-as-new-yorks-ai-safety-law-takes-effect/
  3. Daily Inference. (2026, November 10). Skip an AI safety filing in New York and the fines start at $1 million. https://dailyinference.news/articles/2026-11-10-skip-an-ai-safety-filing-in-new-york-and-the-fines-start-at-1-million/
  4. Google DeepMind. (2026, November 11). AlphaGenome 2: Deciphering how cells read DNA to better understand disease. https://deepmind.google/blog/alphagenome-2-deciphering-how-cells-read-dna-to-better-understand-disease/
  5. IBM. (2026, November 11). IBM Quantum Nighthawk demonstrates new levels of speed and quality, with plans to scale beyond 1,000 qubits in 2027. https://newsroom.ibm.com/2026-11-11-ibm-quantum-nighthawk-demonstrates-new-levels-of-speed-and-quality,-with-plans-to-scale-beyond-1,000-qubits-in-2027
  6. MLQ. (2026, November 11). Google DeepMind releases AlphaGenome 2 to predict how DNA changes affect specific cell types and disease. https://mlq.ai/news/google-deepmind-releases-alphagenome-2-to-predict-how-dna-changes-affect-specific-cell-types-and-disease/
  7. SiliconANGLE. (2026, November 11). IBM Quantum announces upcoming Nighthawk r3 processor designed to execute 10,000 operations. https://siliconangle.com/2026/11/11/ibm-quantum-announces-upcoming-nighthawk-r3-processor-designed-to-execute-10000-operations/
  8. Unite.AI. (2026, November 11). Google DeepMind’s AlphaGenome 2 predicts cell-state variant effects. https://www.unite.ai/google-deepminds-alphagenome-2-predicts-cell-state-variant-effects/
  9. Womack, C. B., Bouabid, S., Sokolov, A., Salunke, P., Flierl, G., Eastham, S. D., & Selin, N. E. (2026). Optimal scenario design for climate emulation. Proceedings of the National Academy of Sciences, 123(46), e2620731123. https://doi.org/10.1073/pnas.2620731123
  10. Zewe, A. (2026, November 12). New research could lead to faster, trustworthy AI models for climate prediction. MIT News. https://news.mit.edu/2026/new-research-could-lead-faster-trustworthy-ai-models-climate-prediction-1112

Similar Posts