{"slug": "autonomous-agent-loops-are-real-karpathy-stripe-and-anthropic-s-dreaming", "title": "Autonomous Agent Loops Are Real: Karpathy, Stripe, and Anthropic's Dreaming", "summary": "Andrej Karpathy's \"autoresearch\" agent ran 700 experiments in two days and surfaced 20 real optimizations for training a small language model, including a missing multiplier in the attention mechanism; applied to a larger model, the fixes cut training time 11%. Shopify's Tobias Lütke reproduced the run on internal data with 37 overnight experiments and a 19% gain, while Stripe's Goose-based agents merge roughly 1,300 pull requests a week behind strict isolation and mandatory human review, and Anthropic documents a \"dreaming\" mechanism in which agents re-read their own session history to extract reusable patterns. The common thread, the source argues, is verification that does not depend on the agent itself: \"A loop with no exit gate and no external check accumulates errors as fast as it accumulates iterations.", "body_md": "Three examples of AI agents running in loops without constant supervision, verified independently rather than taken on a post's word: Andrej Karpathy ran an agent through 700 experiments in two days and pulled 20 real improvements out of it. Stripe has internal agents merging 1,300 pull requests a week, under strict guardrails. Anthropic officially documents a mechanism, dreaming, where the agent re-reads its own history to extract reusable knowledge. A fourth number circulating alongside these, a 42% productivity gain attributed to \"Claude Code's lead engineer,\" has no independent source confirming it.\n\nA regular prompt answers once. A loop re-runs the agent with the previous round's result, until a stop condition is met: a number of iterations, a validated quality bar, a time limit. The difference isn't the model's power, it's the architecture around it: who triggers the loop, who checks the result, and who decides when to stop.\n\nAndrej Karpathy built a system, nicknamed \"autoresearch,\" that proposes a hypothesis, runs an experiment, reads the result, and starts again. In two days, he ran 700 experiments and pulled 20 real optimizations out of it for training a small language model, including a missing multiplier in the attention mechanism that was spreading attention across too many heads at once. Applied to a larger model, those 20 optimizations delivered an 11% cut in training time.\n\nShopify founder Tobias Lütke reproduced the experiment on internal data: 37 experiments overnight, for a 19% gain. Two independent results, which sets this story apart from a one-off anecdote.\n\nAt Stripe, an employee kicks off a development task with an emoji reaction on Slack. An agent, built on a fork of the open source tool Goose, takes it from there: an isolated virtual environment, no internet or production access, preloaded with the necessary code and services in about ten seconds. The result lands as a pull request, never merged without human review.\n\nThe system now handles roughly 1,300 pull requests a week, up 30% in under two weeks from a prior 1,000. The number isn't the point, the architecture that makes it possible is: strict isolation, no direct production access, and a human gate before every merge.\n\nAnthropic officially documents a mechanism called dreaming: an agent that systematically re-reads the history of its own sessions to extract reusable patterns, instead of starting from zero on every task. In a security research demo published in April 2026, nine agents working in parallel spent 800 cumulative hours on a single problem, at a compute cost of about $18,000, and closed 97% of the performance gap on the target task.\n\nAnthropic also states that Claude now writes 80 to 90% of the code for its own internal tools. That figure comes from an official company statement, not a post relaying it.\n\nA fourth item was circulating alongside these three: a knowledge-graph system attributed to \"Claude Code's lead engineer,\" with a claimed 42% productivity gain on day one. No independent source, no technical writeup, no mention on Anthropic's official channels confirms either the number or the attribution. Unlike the three examples above, this one stays a post's number, not a verified fact.\n\nThe three examples that hold up share one trait: verification that doesn't depend on the agent itself. Karpathy reads the experiment results, Stripe requires human review before merging, Anthropic's dreaming mechanism relies on a verifiable history rather than the agent's self-assessment mid-task. A loop with no exit gate and no external check accumulates errors as fast as it accumulates iterations.", "url": "https://wpnews.pro/news/autonomous-agent-loops-are-real-karpathy-stripe-and-anthropic-s-dreaming", "canonical_source": "https://dev.to/dimsbuild/autonomous-agent-loops-are-real-karpathy-stripe-and-anthropics-dreaming-5104", "published_at": "2026-11-19 23:35:07+00:00", "updated_at": "2026-11-19 23:43:05.389567+00:00", "lang": "en", "topics": ["ai-agents", "artificial-intelligence", "large-language-models", "ai-tools"], "entities": ["Andrej Karpathy", "Stripe", "Anthropic", "Shopify", "Tobias Lütke", "Goose", "Claude"], "also_reported_by": [], "takeaway": "Wrapping an agent in a loop that re-feeds each round's result until a stop condition is met, with an external check such as human review or read experiment results, produced 20 optimizations and an 11% training-time cut for Karpathy and about 1,300 reviewed pull requests a week at Stripe.", "recommendation": {"advice": "Gate every agent loop with an external check — human review before merge, read experiment results, or verifiable history — and a defined stop condition, rather than trusting the agent's own self-assessment.", "playbook": "agent-architecture", "playbook_url": "https://wpnews.pro/playbooks/agent-architecture"}, "alternates": {"html": "https://wpnews.pro/news/autonomous-agent-loops-are-real-karpathy-stripe-and-anthropic-s-dreaming", "markdown": "https://wpnews.pro/news/autonomous-agent-loops-are-real-karpathy-stripe-and-anthropic-s-dreaming.md", "text": "https://wpnews.pro/news/autonomous-agent-loops-are-real-karpathy-stripe-and-anthropic-s-dreaming.txt", "jsonld": "https://wpnews.pro/news/autonomous-agent-loops-are-real-karpathy-stripe-and-anthropic-s-dreaming.jsonld"}}