Co-Existence released, uses Anthropic Opus 5.5 to chart 30 months of AI progress
The author’s new book Co‑Existence launched today, looking back at AI development since the early‑2024 release of Co‑Intelligence. Using Anthropic’s Opus 5.5 in Claude Projects, the author let a single “orchestrator” agent spin off additional agents to research, graph, layout, and cite data, turning weeks of work into minutes. The resulting charts illustrate how the AI landscape has shifted from…
Key points
- 180 AI models now surpass GPT‑4’s launch performance, 19 can run on a modern smartphone.
- METR benchmark shows work capacity doubled eight times, from a 4‑minute task in 2024 to 17 hours by April 2026.
- Cost of Google Proof Q&A exam dropped from 30 cents per answer to four hundredths of a cent.
According to the post, 180 AI models now outperform GPT‑4 at its launch, and 19 of those can run on a modern smartphone. The METR Long Horizon Benchmark shows the amount of work an AI can accomplish with a single prompt doubled eight times, growing from a four‑minute task in 2024 to 17 hours of human work by April 2026; some studies suggest the capacity may already exceed 17 weeks. At the same time, the cost of answering a Google Proof Q&A exam fell from 30 cents per answer to four hundredths of a cent. The author argues that these rapid gains are reshaping how people collaborate with AI, moving from chat‑based “centaur” work to autonomous agents that can generate content such as videos with minimal human direction.
Model page: Opus 5.5 →
The story so far
3 episodes →- Co-Existence released, uses Anthropic Opus 5.5 to chart 30 months of AI progressthis story
30 Months on the Exponential (in Charts)
One Useful Thing · 20 October 2026
Loading the full article…
This text was published by One Useful Thing and written by Ethan Mollick. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Researcher says GPT-6 Astra reads Notepad memory in milliseconds without user consent · 1 src
- OpenAI improves code migration for COBOL with GPT-6.1 Sol achieving 97.9% pass rate · 4 src
- MiniMax launches M3.1 Flash model at $0.15 per million input tokens with weights due October 23 · 5 src
- Tsinghua University releases PPMem, a plane-based memory for AI video models · 1 src
- Ornith-ai/Ornith-1.5-122B (Qwen finetune) · 2 src
Comments
via GitHub Discussions