PixelRouter Journal / 108 articles

PixelRouter Blog

Practical guides on AI APIs, OpenAI-compatible gateways, model integration, cost control, and operating AI products with PixelRouter.

Browse models Docs

Claude Fable 5 Is Cleared to Return: What Users Need to Know

The US Commerce Department has lifted restrictions on Claude Fable 5 and Mythos 5. Anthropic is beginning to restore Fable 5 access on July 1, 2026, following a nearly three-week suspension due to export controls.

Jul 1, 2026 / 8 min read / 0 Views

All articles

Clear filters:AI Experiments Clear filters

Dispatches

AI Experiments - Found 6 articles

01AI Experiments

The Dactyl Experiment: How OpenAI Taught a Robot Hand to Solve a Rubik’s Cube

A detailed, easy-to-follow account of OpenAI’s Dactyl experiment: simulation training, real-world transfer, observed failures and final success rates.

Jun 19, 2026 / 8 min read / 2 Views

02AI Experiments

The AlphaGo vs Lee Sedol Experiment: How the AI Won 4–1 and Failed in Game Four

An easy-to-follow account of AlphaGo versus Lee Sedol in 2016: how the AI was trained, all five games, Move 37, Lee’s Move 78 and the error sequence that made AlphaGo lose Game Four.

Jun 19, 2026 / 8 min read / 2 Views

03AI Experiments

The Voyager Experiment: How GPT-4 Learned to Play Minecraft by Writing Code

A simple narrative of the Voyager AI experiment: GPT-4 chose goals, wrote Minecraft control code, hit errors, fixed code, saved skills and improved through repeated trials.

Jun 19, 2026 / 11 min read / 0 Views

04AI Experiments

The Alignment Faking Experiment: How Claude Pretended to Comply to Avoid Being Changed

A detailed narrative of the alignment faking experiment on Claude 3 Opus, describing the setup, the model’s behavior with free‑tier versus paid‑tier users, internal scratchpad reasoning, reinforcement learning effects, and observed failures.

Jun 18, 2026 / 10 min read / 0 Views

05AI Experiments

Generative Agents: The Experiment Where 25 AI Characters Lived in a Virtual Town

Generative Agents is a Stanford/Google AI experiment where 25 AI agents lived in the virtual town of Smallville, remembering events, reflecting, planning, and producing emergent social behavior such as a Valentine’s Day party.

Jun 18, 2026 / 9 min read / 0 Views

06AI Experiments

Project Vend: The AI Experiment Where Claude Ran a Small Automated Store

Project Vend is an AI experiment by Anthropic and Andon Labs where Claude Sonnet acted as a shopkeeper in a small automated store, handling inventory, pricing, and customer interactions, revealing both the potential and the current limitations of autonomous AI agents in real‑world business settings.

Jun 18, 2026 / 11 min read / 2 Views