# Autoresearch by Karpathy Hacker Approved page: https://hackerapproved.com/p/autoresearch/ 25 people recommended Autoresearch by Karpathy in Hacker News comments, 9 had mixed feelings and 13 criticized it, November 2025 to October 2026. It was recommended in 6 of 12 months, in 17 threads. It does not meet the Hacker Approved badge rule. **Verdict:** Recommended as a free, simple way to let an agent run experiments against a clear metric; critics say it mostly tunes parameters and classical optimization often does as well. Summary written by Claude from 48 comments. It was written before the November 2026 recount, and 44 of those comments still count. ## Why people recommend it - Works well with a clear success metric (3 comments in the sample): "Shows the power of Autoresearch when you're actually able to build a strict set of tests that clearly delineate successful outcomes." ([source](https://news.ycombinator.com/item?id=48683692)) "the ability to say: "optimize this benchmark and commit only if you pass these tests" is nice" ([source](https://news.ycombinator.com/item?id=47494169)) "extremely powerful when applied to problems that generalize well to it" ([source](https://news.ycombinator.com/item?id=48713750)) - Speeds up code and test suites (3 comments in the sample): "autoresearch managed to improve the harness performance on my 'time-to-Realworld' benchmark" ([source](https://news.ycombinator.com/item?id=47551350)) "It's nice being able to just run /autoresearch and speed up your test suites" ([source](https://news.ycombinator.com/item?id=48205114)) "Closest thing to “Make this app fast!” we have now." ([source](https://news.ycombinator.com/item?id=49397440)) - Automates parameter sweeps and experiments (2 comments in the sample): "I can now very lazily set up large parameter sweeps and visualization scripts to really probe things." ([source](https://news.ycombinator.com/item?id=47500464)) "This automated run-and-discover process is far beyond what’s possible with hyperparam search." ([source](https://news.ycombinator.com/item?id=47446644)) - Cheap, accessible way to experiment (2 comments in the sample): "autoresearch feels more accessible that the listed solutions." ([source](https://news.ycombinator.com/item?id=47500222)) "Perfect for experimenting with AI research on the cheap!" ([source](https://news.ycombinator.com/item?id=47385923)) ## What people warn about - Not groundbreaking; mostly hype (4 comments in the sample): "nothing particularly groundbreaking, but has helped devs expand their understanding of the utility that these models can provide" ([source](https://news.ycombinator.com/item?id=48194695)) "Karpathy writes a really basic loop and it's now a kind of AI miracle prompting tons of grifters, copy-cats, weird hype." ([source](https://news.ycombinator.com/item?id=47444889)) "Neither of them were even original, people have been playing around with those ideas for well over a year." ([source](https://news.ycombinator.com/item?id=48197614)) - Mostly hyperparameter tuning, not novel ideas (3 comments in the sample): "But what it ends up doing is tuning parameters manually, only one / a few at a time, extremely inefficient and unlikely to be optimal." ([source](https://news.ycombinator.com/item?id=47503701)) "basically everything is just hyperparameter tuning. Which is nice, but likely not worth the $$$ spent on the tokens" ([source](https://news.ycombinator.com/item?id=47493988)) "it was just tuning parameters mostly, not really trying different novel approaches" ([source](https://news.ycombinator.com/item?id=48200873)) - Classical optimization does as well or better (3 comments in the sample): "the vast majority of the changes that auto research actually made would have been found much faster with BO" ([source](https://news.ycombinator.com/item?id=47449675)) "TPE crushes Karpathy’s autoresearch and it is neck and neck with the method in this paper" ([source](https://news.ycombinator.com/item?id=48466105)) "one experiment showing that classical black-box optimization did better than Autoresearch" ([source](https://news.ycombinator.com/item?id=47494263)) - Random changes and unreadable code (2 comments in the sample): "they tend to randomly change things, which is a poor approach" ([source](https://news.ycombinator.com/item?id=47494761)) "Half the tests aren't passing and the code is a lot less readable, but the core idea is cool" ([source](https://news.ycombinator.com/item?id=47373519)) ## Other research agents people recommend - [Gemini Deep Research](https://hackerapproved.com/p/gemini-deep-research/): 20 recommend, 3 mixed, 17 criticize - [OpenAI Deep Research](https://hackerapproved.com/p/openai-deep-research/): 7 recommend, 5 mixed, 13 criticize - [Tongyi DeepResearch](https://hackerapproved.com/p/tongyi-deepresearch/): 7 recommend, 0 mixed, 0 criticize - [Kagi Research](https://hackerapproved.com/p/kagi-research/): 6 recommend, 2 mixed, 0 criticize ## The badge A product earns the Hacker Approved badge when at least 10 people recommend it, recommenders outnumber critics at least 2 to 1 (mixed feelings count toward neither side), it was recommended in at least 4 of the 12 months and in at least 5 different threads, and no single thread supplies more than half of the people who recommend it. Health and money products are never badged. ## Cite as Hacker Approved, "Autoresearch by Karpathy", https://hackerapproved.com/p/autoresearch/ (25 recommend, 9 mixed, 13 criticize; Hacker News comments, November 2025 to October 2026). Counts updated 2 Nov 2026, with data through October 31, 2026.