Observed Signal · Mar 20, 2026 · Technical Release · Source: Aakash Gupta · Impact: 3/5 · Sentiment: Positive
Unlocking Autoresearch: A Game-Changer for Product Managers
Andrej Karpathy released an open-source project called "autoresearch," an agentic loop that autonomously iterates code changes, runs short experiments, evaluates a numeric metric, and commits improvements. The system has drawn large interest (noted as ~42,000 GitHub stars) and produced measurable gains in small‑model training and real-world codebases: Karpathy’s agent found multiple improvements that transferred to larger models, and Shopify CEO Tobi Lutke reported a 53% faster parse+render for Shopify’s Liquid templating engine after automated commits. Product manager Aakash Gupta published a practical guide for PMs explaining how to apply the pattern to prompts, skills, and templates, including setup steps, six use cases, eval templates, and a toolkit. The pattern requires a clear numeric metric, an automated evaluator, and a single editable file; Gupta recommends tools such as Claude Code or other coding agents to run the loop overnight.
An open-source technical release demonstrating autonomous, closed-loop optimization for code and prompts can accelerate experimentation and product improvements—relevant for PMs and engineering teams, and shown to produce substantive performance gains in e-commerce and model training.
Track Shopify Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Andrej Karpathy published an open-source "autoresearch" repo that automates propose-change → run short experiment → evaluate → commit loop.
- The autoresearch repo has been widely noticed (reported ~42,000 GitHub stars in the article).
- Karpathy’s agent discovered multiple improvements that stacked and transferred to larger models, yielding an 11% speedup in one reported case.
- Shopify CEO Tobi Lutke applied an autonomous researcher to Shopify’s Liquid templating engine and reported 53% faster rendering and 61% fewer object allocations from 93 automated commits.
- Product manager Aakash Gupta published a PM-focused guide and toolkit showing how to set up autoresearch for prompts/skills using tools like Claude Code.
Connected Companies & Entities
1 Entity mappedOntology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Karpathy's Autoresearch Spurs AutoBeta Agent Experiments
Andrej Karpathy released ~600 lines of Python implementing 'autoresearch', an autonomous experimental loop that runs hypothesis-test-score-iterate cycles under human-set objectives and constraints. In Karpathy’s initial run it trained a GPT-2–level model in two days, achieving an 11% speed improvement and finding 20 genuine improvements. Shopify CEO Toby Lütke applied autoresearch to Shopify’s internal model (qmd), running 37 experiments overnight and producing a 0.8-billion-parameter model that outscored a prior 1.6-billion-parameter version by 19%. The Exponential View author adapted the loop for general knowledge work as
Karpathy's Autoresearch Enables Automated Creative Optimization
The newsletter deep-dive explains Andrej Karpathy’s newly open-sourced “autoresearch” repo — an automated loop that runs large numbers of variations overnight to improve prompts, code, copy and other measurable outputs — and shows how marketers can apply it to ad copy, email sequences, landing pages, video scripts and job posts. The piece also summarizes major industry moves: Google announced Gemini-powered Ask Maps and Immersive Navigation as the biggest Maps AI upgrade in over a decade; Anthropic added features like Dispatch/Cowork and in-conversation visualizations; and examples from practitioners (Tobi Lutke, Single Grain, MindStudio) demonstrate large, low-cost gains when applying autoresearch to real systems.
Agentic Engineering: PMs Review Artifacts, Not Code
A product manager describes a shift in PM workflows driven by AI coding agents: instead of reading code, PMs should maintain and review the artifact layer (strategy files, agent contracts, CLAUDE.md, tests, evals) that steers agents. The author shipped three projects (PM Brain, Claude Usage for VS Code, Grok Build), ran 800+ tests and LLM-based evals, and published an "AI Shipping Artifact Prompt Pack" (artifact prompts + audit commands). Key practices include a single source-of-truth document for agents, triage rules that combine soft steering with mechanical guardrails, cross-model review to catch blind spots, and converting failures into permanent tests or policies. The piece argues prototypes and agent-driven builds now often precede full alignment, so artifact maintenance and selective human pushback are the primary PM responsibilities when working with agentic systems.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.