AI intelligence without the noise
Start with today’s source-backed briefing, then explore plain-English explainers, model comparisons, prices, markets, industry moves, policy, research and AI safety.
Today in 60 seconds
Anthropic expanded general availability of Claude Fable 5.1 for enterprise multi-step reasoning, agents, and codebase migrations, featuring 1M context and cutting prompt cache-read costs from $1.00 to $0.25 per million tokens.
Reduces the operational cost of multi-turn software agents by up to 90%, transforming long-running autonomous developer workflows from an expensive novelty into an economically sustainable enterprise tool.
Frontier models (Claude Opus 5 at 96.0%, Claude Fable 5.1 at 95.0%, and OpenAI Astra at 94.2%) saturated SWE-bench Verified, prompting evaluation consortiums to formalize a transition to private, polyglot SWE-bench Pro suites.
Data contamination across open-source GitHub issues has rendered SWE-bench Verified incapable of distinguishing true software engineering capability from pre-training memory; early results on SWE-bench Pro drop resolution rates to 48%–56%.
Today’s explainer
Featured explainer · August 22
A plain-English guide to calculating the real savings from OpenAI’s temporary GPT-5.6 Sol price cut, including input versus output tokens, workload mix, caching, tools, retries and the three-month expiry.
Recent explainer · August 20
OpenAI says Private Safety Processing can look for dangerous patterns across related interactions while keeping eligible customer content under Zero Data Retention protections. Here is what that means, how the proposed architecture works, what it does not yet prove, and which questions enterprise buyers should still ask.
Recent explainer · August 9
How AI factories differ from general-purpose data centers, why GPUs are only one part of the system, and why power, networking, cooling and utilization matter.
Recent explainer · August 9
Why many slightly different forecasts can be more useful than one line, what spread and calibration mean, and how AI changes probabilistic forecasting.
Recent explainer · August 9
OpenAI’s High and Critical cybersecurity thresholds, the evaluations behind them, why autonomy matters, and why “cannot rule out Critical” is not a confirmed classification.
Recent explainer · August 9
What Medium, High and Extra High reasoning change, why more inference compute can help difficult tasks, and why more reasoning does not guarantee correctness.
Tuesday complete daily edition · Tuesday, September 8, 2026
Today’s AI intelligence
Anthropic rolls out Claude Fable 5.1 with slashed prompt-cache rates alongside restricted Mythos 5.1 cyber models; frontier labs abandon saturated public code benchmarks for SWE-bench Pro; and hyperscalers retool infrastructure around rack-level agentic throughput.
In plain English: Frontier artificial intelligence has reached a key transition point where raw single-turn reasoning records matter less than whether autonomous agents are economically affordable and safely contained. Anthropic expanded general availability of Claude Fable 5.1 today, cutting the cost to read cached memory by 75%—a move that drops the cost of long multi-step agent workflows by up to 90%. Meanwhile, with leading models now clustering above 94% on SWE-bench Verified, researchers are migrating to private, contamination-resistant evaluation suites. In datacenters, cloud providers are committing billions to liquid-cooled, rack-scale hardware designed specifically to handle continuous memory caching and sub-200 millisecond agent responses.
Three developments
Frontier Agentic Models & Economics
Evaluation & Benchmark Integrity
Datacenter Silicon & Infrastructure
Intelligence databases
Current model families, providers, access, context and published prices.
Open models →Benchmarks6 categoriesSeparate leaderboards for reasoning, mathematics, coding, agents, long context and documents.
Open benchmarks →Prices5 comparedPublished API prices and a dated history of verified changes.
Open prices →Markets3 trackedA concise view of selected United States AI-related market closes.
Open markets →Industry5 developmentsCompany, infrastructure and product developments with practical consequences.
Open industry →Research & policy3 monitoredResearch, safety and policy changes separated from commercial product news.
Open research & policy →How AI Works
A plain-English explanation of how prompt caching works at the GPU memory layer, why stateless multi-turn agent loops cost 10x more, and how 75% cache discounts make autonomous software engineering economically viable.
Trust and verification
Explore AIUpdateWatch
News, models, costs, industry, research and practical explanations—organized so you can go directly to the level of detail you need.
What changed today, why it matters and what to watch next.
Plain-English explanations of how AI works and how people use it.
Frontier, open and local models, plus the evidence used to compare them.
What AI costs to use and what the commercial AI market is doing.
Companies, infrastructure, regulation, geopolitics and the institutions shaping AI.
New research, reliability, security, privacy and frontier-model safety.