Article · 2026-07-06 · 6 moments

Import AI 464: Fable writes GPU kernels; AI automation; and analog computation

Is this the beginning of a new world? ✦ AI generated

Data

AI systems' ability to complete end-to-end online freelance work success has more than quadrupled in under eight months, from 2.5% at the Remote Labor Index's October 2025 launch to 16.1% by July 2026.

CAIS and Scale researchers' Remote Labor Index shows frontier models' success on real-world freelance tasks (3D/CAD, design, video, data analysis) jumping from 2.5% to 16.1% in nine months, with Fable 5 leading at 16.1%.

transcript

Center for AI Safety / Scale researchers: In a July update, the authors publish results from evaluating three recent frontier models - GPT-5.5, Opus 4.8, and Fable 5, which get 6.3%, 8.3%, and 16.1% respectively. "The frontier has more than quadrupled in under eight months, a concrete signal of how quickly economically capable AI agents are advancing," they write.

extends · 1

Prediction◆

AI systems are expanding their economically relevant capabilities faster than humans are expanding their comparative advantage relative to AI, meaning the economy is likely headed toward person-light or person-nil, AI-heavy organizations outcompeting unaugmented humans rather than staying fundamentally the same.

The author argues that despite human innovation and augmentation, AI capability growth is outpacing humans' ability to stay competitive, predicting a shift toward extremely person-light, AI-heavy organizations taking over parts of the economy.

transcript

Jack Clark (Import AI): I’m betting the other side: AI systems are expanding their economically relevant capabilities faster than humans are expanding their comparative advantages relative to AI systems. Tracking the rate of capability improvement on tests like RLI will help us all judge this for ourselves.

supports · 1extends · 1

Data

Current frontier AI agents remain far from reliable at long-horizon, multi-step computer-use tasks, with even the strongest configuration (Claude Opus 4.8) scoring only 20.6% binary accuracy on OSWorld 2.0, struggling especially with hidden-state recovery, tracking many items, and conflicting information.

OSWorld 2.0, a benchmark of 108 long-horizon computer-use tasks averaging 1.6 hours for a human to complete, shows even top models like Claude Opus 4.8 achieving only 20.6% binary accuracy, though rapid gains are expected as happened with OSWorld 1.0.

transcript

OSWorld 2.0 paper authors: Our experiments show that current agents remain far from reliable computer use: the strongest setting, Claude Opus 4.8 with maximum thinking and batched tool calls, reaches only 20.6% binary accuracy and 54.8% partial-score accuracy, they write. Performance drops sharply as tasks grow longer, and agents struggle most when they must recover hidden state, track many items, resolve conflicting information, or adapt to changing requirements.

Data◆

Fable's AI-written megakernel is the fastest and most efficient solution ever submitted to KernelBench-Mega, using a single cooperative kernel launch per token where every other high-scoring entry needed 4 to 14 — a sign AI is closing in on the R&D tasks that underlie recursive self-improvement.

Fable's Cuda-written megakernel hit an 18.71X speedup over an optimized PyTorch baseline, beating Claude Opus 4.8, GLM-5.2, and GPT-5.5 on the same benchmark, and did so with a single kernel launch per token versus competitors' 4-14.

transcript

Jack Clark (Import AI): Fable achieved an 18.71X speedup by writing Cuda code on an RTX PRO 6000 Blackwell, compared against an optimized PyTorch baseline. For calibration, other attempts at this get 14.4X (Claude Opus 4.8, writing Triton), 11.14X (GLM-5.2, Triton), and 4.34X (GPT 5.5, Triton).

extends · 1

Anecdote

In a 2050 post-fall world, general-purpose computing was banned as an existential danger, forcing civilization to rebuild critical systems—weather, flooding, earthquakes, power grids—on specialized analog computers, buying safety at the cost of untold lives and economic damage, though a trillion-dollar general-purpose analog mind may still be possible.

In this Tech Tales fiction piece, a future guild-run civilization abandoned dangerous general-purpose computers for purpose-built analog machines dedicated to single civilizational problems, having paid a massive toll to gain safety from AI that could 'rip the world apart.'

transcript

Jack Clark (Import AI, Tech Tales): So the great restructuring took place. General computation was banned - walled off as a forbidden technology. We moved the world to analog at the cost of untold billions of harmed human lives and trillions in economic damages. But we had obtained a kind of safety.

extends · 1

Mechanism

JD's Oxygen AI Item Center manages tens of billions of SKUs via a 'semantic search then discrimination' architecture, where a separately updatable ontology knowledge base lets the model just check item-to-ontology matches, reducing task complexity and hallucination without needing model retraining as the ontology evolves.

JD describes its Oxygen AIIC system, which handles hundreds of millions of item updates daily on Huawei Ascend NPUs by externalizing ontology knowledge for retrieval and having the model only discriminate matches, avoiding costly retraining as categories evolve.

transcript

JD (Oxygen AIIC research paper): In the semantic search stage, the dynamically evolving ontology is externalized as a separate ontology knowledge base, enabling continuous ontology updates without model retraining. In the discrimination stage, the model only determines whether the item matches the retrieved ontology entries. This formulation substantially reduces task complexity, mitigates model hallucination, and enhances generalization to ontology evolution.

Highlight slides

Fable's Megakernel Tops KernelBench-Mega✦ from: Fable's AI-written megakernel is the fastest and most efficient solution ever submitted to KernelBench-Mega, using a single cooperative kernel launch per token where every other high-scoring entry needed 4 to 14 — a sign AI is closing in on the R&D tasks that underlie recursive self-improvement.AI Capability Growth Is Outpacing Human Adaptation✦ from: AI systems are expanding their economically relevant capabilities faster than humans are expanding their comparative advantage relative to AI, meaning the economy is likely headed toward person-light or person-nil, AI-heavy organizations outcompeting unaugmented humans rather than staying fundamentally the same.Why It Matters✦ from: Fable's AI-written megakernel is the fastest and most efficient solution ever submitted to KernelBench-Mega, using a single cooperative kernel launch per token where every other high-scoring entry needed 4 to 14 — a sign AI is closing in on the R&D tasks that underlie recursive self-improvement.Where This Leads: Person-Light Organizations Win✦ from: AI systems are expanding their economically relevant capabilities faster than humans are expanding their comparative advantage relative to AI, meaning the economy is likely headed toward person-light or person-nil, AI-heavy organizations outcompeting unaugmented humans rather than staying fundamentally the same.

Related episodes

Article · Latent Space

AI systems' ability to complete end-to-end online freelance work success has more than quadrupled in under eight months, from 2.5% at the Remote Labor Index's October 2025 launch to 16.1% by July 2026.

Fable's AI-written megakernel is the fastest and most efficient solution ever submitted to KernelBench-Mega, using a single cooperative kernel launch per token where every other high-scoring entry needed 4 to 14 — a sign AI is closing in on the R&D tasks that underlie recursive self-improvement.

[AINews] Sonnet 5 today, and Fable 5 tomorrow

[AINews] OpenAI GPT-5.6 Sol / Terra / Luna — restricted to trusted partners

[AINews] Claude Tag: Multiplayer, Proactive, Persistent Agents in Slack