Run step
Step 4: tool results
search_similar_claims
Input
{
"limit": 12,
"query": "AI systems autonomously conduct AI research and train successor models; automated AI R&D timeline"
}{
"query": "AI systems autonomously conduct AI research and train successor models; automated AI R&D timeline",
"results": [
{
"id": "e083ea2a-d43e-4156-8cca-395ac22ef6b5",
"kind": "quantity",
"text": "Calendar year when AI systems first automate 99% of AI research and development tasks",
"claim_type": "empirical_derived",
"state": "active",
"similarity_score": 0.6983665925918242,
"importance": 0.8,
"assessment_status": null,
"assessment_confidence": null
},
{
"id": "cb0cfaf6-af14-493f-8fc3-0d571ecb3fc9",
"kind": "quantity",
"text": "Calendar year by which AI systems first perform most AI research and development tasks",
"claim_type": "empirical_derived",
"state": "active",
"similarity_score": 0.6953274530586023,
"importance": 0.75,
"assessment_status": null,
"assessment_confidence": null
},
{
"id": "886f84f4-62d6-4899-afd7-8603067caf49",
"kind": "quantity",
"text": "AI software-efficiency doubling time immediately after full automation of AI R&D, in months",
"claim_type": "empirical_derived",
"state": "active",
"similarity_score": 0.6933188724073232,
"importance": 0.45,
"assessment_status": null,
"assessment_confidence": null
},
{
"id": "aa36d48a-c988-4991-a701-50102794ef74",
"kind": "quantity",
"text": "Calendar year in which AI systems first fully automate AI research and development",
"claim_type": "empirical_derived",
"state": "active",
"similarity_score": 0.6912623511179321,
"importance": 0.85,
"assessment_status": null,
"assessment_confidence": null
},
{
"id": "5e4c99e3-f697-45fd-b79e-d75cb33a5732",
"kind": "claim",
"text": "By the end of 2028, a frontier AI model will autonomously train a successor version of itself.",
"claim_type": "empirical_verifiable",
"state": "active",
"similarity_score": 0.6848675864243504,
"importance": 0.85,
"assessment_status": null,
"assessment_confidence": null
},
{
"id": "55f89371-8a43-43c1-8dd7-932c350ca7fb",
"kind": "claim",
"text": "By the end of 2027, a frontier AI model will autonomously train a successor version of itself.",
"claim_type": "empirical_verifiable",
"state": "active",
"similarity_score": 0.6835670902145949,
"importance": 0.75,
"assessment_status": null,
"assessment_confidence": null
},
{
"id": "44caf44c-3dbc-4084-944f-687f2fb28510",
"kind": "claim",
"text": "Anthropic models available by July 15, 2026, could fully automate AI R&D and R&D in other key domains.",
"claim_type": "empirical_derived",
"state": "active",
"similarity_score": 0.6741482895090697,
"importance": 0.6,
"assessment_status": null,
"assessment_confidence": null
},
{
"id": "f5a654f9-9ef3-4144-9869-6bff1c6d9631",
"kind": "claim",
"text": "AI systems are improving at choosing next steps in AI research.",
"claim_type": "empirical_derived",
"state": "active",
"similarity_score": 0.6681576925123482,
"importance": 0.55,
"assessment_status": "supported",
"assessment_confidence": 0.75
},
{
"id": "e68a817d-3271-4cb1-88c1-a7d7a2443a83",
"kind": "claim",
"text": "AI automation of AI research could trigger rapid recursive growth in AI capabilities.",
"claim_type": "causal",
"state": "active",
"similarity_score": 0.6623946512305932,
"importance": 0.75,
"assessment_status": "supported",
"assessment_confidence": 0.62
},
{
"id": "f07e018c-7df9-4db6-972c-d7feb111e699",
"kind": "claim",
"text": "AI systems will become capable of fully automating frontier AI research.",
"claim_type": "empirical_derived",
"state": "active",
"similarity_score": 0.6563641916538563,
"importance": 0.75,
"assessment_status": "supported",
"assessment_confidence": 0.55
},
{
"id": "a3a7a95b-f3fb-4cf8-a605-cbe3bee834db",
"kind": "claim",
"text": "AI systems perform a large and rapidly growing share of AI research and development at frontier labs.",
"claim_type": "empirical_derived",
"state": "active",
"similarity_score": 0.6517054132045424,
"importance": 0.6,
"assessment_status": null,
"assessment_confidence": null
},
{
"id": "3699eb33-020b-43bd-8da6-469f19575f06",
"kind": "claim",
"text": "Partial automation of AI R&D tasks will substantially speed AI software progress before full automation.",
"claim_type": "causal",
"state": "active",
"similarity_score": 0.650590260028532,
"importance": 0.7,
"assessment_status": "supported",
"assessment_confidence": 0.65
}
],
"count": 12
}provenance_read_source
Input
{
"max_chars": 8000,
"source_id": "87406e11-4df5-49b3-8558-11e864316894"
}{"success":true,"source":{"id":"87406e11-4df5-49b3-8558-11e864316894","url":"https://importai.substack.com/p/import-ai-455-automating-ai-research","title":"Import AI 455: AI systems are about to start building themselves.","source_type":"blog_post"},"origin":"stored","total_chars":30727,"offset":0,"content":"Subscribe Sign in\n\nImport AI 455: AI systems are about to start building themselves.\n\nThe first step towards recursive self improvement\n\nJack Clark\n\nMay 04, 2026\n\n351\n\n54\n\n67\n\nShare\n\nWelcome to Import AI, a newsletter about AI research. Import AI runs on arXiv and feedback from readers. If you’d like to support this, please subscribe.\n\nSubscribe\n\nAI systems are about to start building themselves. What does that mean?\n\nI’m writing this post because when I look at all the publicly available information I reluctantly come to the view that there’s a likely chance (60%+) that no-human-involved AI R&D - an AI system powerful enough that it could plausibly autonomously build its own successor - happens by the end of 2028.\nThis is a big deal.\nI don’t know how to wrap my head around it.\nIt’s a reluctant view because the implications are so large that I feel dwarfed by them, and I’m not sure society is ready for the kinds of changes implied by achieving automated AI R&D.\nI now believe we are living in the time that AI research will be end-to-end automated. If that happens, we will cross a Rubicon into a nearly-impossible-to-forecast future. More on this later.\n\nThe purpose of this essay is to enumerate why I think the takeoff towards fully automated AI R&D is happening. I’ll discuss some of the consequences of this, but mostly I expect to spend the majority of this essay discussing the evidence for this belief, and will spend most of 2026 working through the implications.\n\nIn terms of timing, I don’t expect this to happen in 2026. But I think we could see an example of a “model end-to-end trains it successor” within a year or two - certainly a proof-of-concept at the non-frontier model stage, though frontier models may be harder (they’re a lot more expensive and are the product of a lot of humans working extremely hard).\nMy reasoning for this stems primarily from public information: papers on arXiv, bioRxiv, and NBER, as well as observing the products being deployed into the world by the frontier companies. From this data I arrive at the conclusion that all the pieces are in place for automating the production of today’s AI systems - the engineering components of AI development. And if scaling trends continue, we should prepare for models to get creative enough that they may be able to substitute for human researchers at having creative ideas for novel research paths, thus pushing forward the frontier themselves, as well as refining what is already known.\n\nUpfront caveat\nFor much of this piece I’m going to try to assemble a mosaic view of AI progress out of things that have happened with many individual benchmarks. As anyone who studies benchmarks knows, all benchmarks have some idiosyncratic flaws. The important thing to me is the aggregate trend which emerges through looking at all of these datapoints together, and you should assume that I am aware of the drawbacks of each individual datapoint.\n\nNow, let’s go through some of the evidence together.\n\nThe coding singularity - capabilities over time:\nAI systems are instantiated via software and software is made out of code.\n\nAI systems have revolutionized the production of code. This has happened due to two related trends: AI systems have gotten better at writing complicated real-world code, and AI systems have gotten much better at chaining together many linear coding tasks (e.g, writing code, then testing it) independent of human oversight.\n\nTwo things that exemplify this trend are SWE-Bench and the METR time horizons plot.\n\nSolving real-world software engineering problems:\nSWE-Bench is a widely used coding test which evaluates how well AI systems can solve real world GitHub issues. When SWE-Bench launched in late 2023 the best score at the time was Claude 2 which had an overall success rate of ~2%. Claude Mythos Preview gets 93.9%, effectively saturating the benchmark. (All benchmarks have some amount of noise inherent to them, so there’s usually a point where you score high enough that you are running into the limitations of the benchmark itself rather than your method - for instance, about 6% of the labels in the ImageNet validation set are wrong or ambiguous).\nSWE-Bench is a reliable proxy for the general issue of coding competency and the impact of AI on software engineering. The vast majority of people I meet at frontier labs and around Silicon Valley now code entirely through AI systems. Increasingly, they use AI systems to write the tests and check the code as well. In other words, AI systems have gotten good enough to automate a major component of AI R&D, speeding up all the humans that work on it.\n\nMeasuring an AI system’s ability to complete tasks that take people a long time:\nMETR makes a plot that tells us about the complexity of tasks AIs can complete, measured by how many hours a skilled human would take to do them. The key measure here is one which tells you the rough time horizon over which AI systems can be 50% reliable at a basket of tasks.\nHere, progress has been extremely striking: In 2022, GPT 3.5 could do tasks that might take a person about ~30 seconds. In 2023, this rose to 4 minutes with GPT-4. In 2024, this rose to 40 minutes (o1). In 2025, it reached ~6 hours (GPT 5.2 (High)). In 2026, it has already risen to ~12 hours (Opus 4.6). Ajeya Cotra, a longtime AI forecaster who works at METR, thinks it isn’t unreasonable to expect AI systems to do tasks that take ~100 hours by the end of 2026 ( #448 ).\nThis significant rise in the length of time that AI systems can work independently correlates neatly with the explosion in agentic coding tools - this is the productization of AI systems which do work on behalf of people, acting independently for significant periods of time.\nIt also loops back to AI R&D, where if you look closely at the work of many AI researchers, a lot of their tasks boil down into things that might take a person a few hours to do - cleaning data, reading data, launching experiments, etc. All of this kind of work now sits inside the time horizon scope of modern systems.\n\nThe more skilled AI systems get and the better they get at working independently of us, the more they can help automate chunks of AI R&D\nKey ingredients in delegation are a) confidence in the skills of the person, and b) confidence in their ability to work independently of you in a way that is aligned with your intentions.\nWhen we look at the competency of AI at coding, it seems that AI systems are getting far more skilled and also able to work independently of people for longer and longer periods before needing re-calibration.\nThis correlates with what we see around us - engineers and researchers are now delegating larger and larger chunks of their work to AI systems, and as capabilities rise, so too does the complexity and importance of the work being delegated.\n\nAI is getting good at core science skills essential to AI R&D\nThink about modern science - a huge amount of it is about specifying a direction where you want to generate some empirical information, running experiments to generate that information, then sanity-checking the results of the experiment. The combination of advances in coding over time combined with the general world modeling capabilities of LLMs has yielded tools that are already helping to speed up human scientists and partially automate aspects of R&D broadly.\n\nHere, we can look at the rate of AI progress in a few key scientific skills which are inherent to AI research itself: Replicating research results, chaining together machine learning techniques and other approaches to solve technical problems, and optimizing AI systems themselves.\n\nImplementing entire scientific papers and doing the experiments:\nOne core job of AI research is reading scientific papers and reproducing their results. Here, there has been dramatic progress on a wide range of benchmarks.\n\nOne good example is CORE-Bench , the Computational Reproducibility Agent Benchmark. This benchmark challenges AI systems to","truncated":true,"next_offset":8000,"facts":{"authors":["Jack Clark"],"publisher":null,"published_date":null,"doi":null,"archived_url":"https://web.archive.org/web/20261112104114/https://importai.substack.com/p/import-ai-455-automating-ai-research","events":[],"versions":[],"copy_of":[]},"note":"More text follows; call again with offset 8000 to continue."}