Prediction Open AI assessment confidence: 85% certainty 3/5 debate potential 2/5

Altman expects coding AI to execute multi-day tasks by 2026

Sam Altman · All things AI w @altcap @sama & @satyanadella. A Halloween Special. 🎃🔥BG2 w/ Brad Gerstner · Bg2 Pod · Oct 31, 2025 · at 28:15

OpenAI CEO Sam Altman discusses the expected evolution of coding models and agents over the next year on the BG2 Podcast.

0:00 / 0:13exact quote · 13.2s

▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark

“Codex has been a very cool thing to watch this year, and as these go from multi-hour tasks to multi-day tasks, which I expect to happen next year, what people be able to do to create Software at an unprecedented rate.”

quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →

The claim, plainly

Sam Altman expects coding AI agents to advance from handling multi-hour tasks to executing multi-day tasks by 2026.

a quote alone can be ambiguous, so this restates it with the references and context of the conversation around it resolved. AI-written; the tape below the player is the source.

Assessment

On October 31, 2025, Sam Altman predicted that AI coding agents would advance from multi-hour to multi-day tasks 'next year' (2026). As of September 5, 2026, approximately four months remain before the stated 2026-12-31 horizon closes. Current benchmark evaluations (such as SWE-Marathon with <30% resolution and SWE-EVO at 22.9-25.0%) and commercial reports show that while agents routinely handle multi-hour tasks, fully autonomous execution of release-scale multi-day tasks remains constrained by compounding errors, context drift, and loop failures. Because the stated deadline has not yet arrived and frontier developments remain active, the claim cannot be resolved definitively until the conclusion of 2026.

SWE-Marathon (June 2026) showed leading frontier coding agents resolved fewer than 30% of ultra-long-horizon tasks due to compounding errors and context drift.
SWE-EVO (August 2026) evaluated release-sized software evolution tasks, finding top frontier models achieved only 22.9% to 25.0% success.
Commercial coding agents as of August 2026 routinely execute multi-hour tasks and hundreds of tool turns asynchronously, but autonomous multi-day execution without human intervention remains constrained by reliability bottlenecks.
The stated prediction horizon of 2026-12-31 has not yet elapsed as of the assessment date of 2026-09-05.

Checked against

[1] localaimaster.com [2] morphllm.com [3] openreview.net [4] icml.cc [5] arxiv.org [6] arxiv.org [7] openreview.net [8] medium.com

long horizon software engineering AI agent multi-day SWE-bench 2026 SWE bench verified leaderboard multi day long horizon software engineering 2025 2026 autonomous AI coding agents multi day tasks 2025 2026 Devin OpenAI Anthropic AI coding agents capable of multi day tasks software engineering 2026

assessed as of 2026-09-05

This assessment is an AI reading of a public statement, made against the cited references with the confidence shown. It can be wrong. The quote, the tape and the references are all on this page; if something is off, use report an error below. full disclaimer →

⏳ prediction date: 2026-12-31; checked again automatically on the next resolution wave

Mentioned

Codex OpenAI

⚑ Spot an error? Report it

More from Sam Altman

Prediction Not checkable as stated

Altman: An AI compute infrastructure glut is guaranteed to happen

“There will come a glut for sure. And whether that's like in two to three years or five to six, Satya and I can't tell you, but like it's gonna happen at some point, probably several points along the way.”

Sam Altman Oct 31, 2025 ▶ 19:29 All things AI w @altcap @sama & @satyanadella. A Halloween Special. 🎃🔥BG2 w/ Brad Gerstner · Bg2 Pod

Assertion Partly supported

Altman: OpenAI is generating well over $13B in revenue

“We're doing well more revenue than that.”

Sam Altman Oct 31, 2025 ▶ 12:35 All things AI w @altcap @sama & @satyanadella. A Halloween Special. 🎃🔥BG2 w/ Brad Gerstner · Bg2 Pod

Prediction Not checkable as stated

Altman: OpenAI will become a major AI cloud and consumer device business

“And that not only will Chachapiti keep growing, but we will be able to become one of the important AI clouds that our consumer device business will be a significant and important thing that AI that can automate science will create huge value.”

Sam Altman Oct 31, 2025 ▶ 13:12 All things AI w @altcap @sama & @satyanadella. A Halloween Special. 🎃🔥BG2 w/ Brad Gerstner · Bg2 Pod

Prediction Open · timeframe Oct 2030

Altman: OpenAI will eventually build low-power devices running GPT-6 locally

“Someday we will make an incredible consumer device that can run a GPT-Five or GPT-Six capable model completely locally at a low power draw.”

Sam Altman Oct 31, 2025 ▶ 21:15 All things AI w @altcap @sama & @satyanadella. A Halloween Special. 🎃🔥BG2 w/ Brad Gerstner · Bg2 Pod

Prediction Held up

Altman predicts AI will make a novel scientific discovery by 2026

“I hope for very small scientific discoveries in 20, 26, but if we can get those very small ones, we'll get bigger ones in future years. That's a really crazy thing to say is that like AI is going to make a novel scientific discovery in 20, 26, even a very smal…”

Sam Altman Oct 31, 2025 ▶ 28:41 All things AI w @altcap @sama & @satyanadella. A Halloween Special. 🎃🔥BG2 w/ Brad Gerstner · Bg2 Pod

Prediction Open · timeframe Oct 2030

Altman predicts OpenAI's nonprofit will become the largest nonprofit ever

“And also as Satya said, create what we believe will be the largest nonprofit ever.”

Sam Altman Oct 31, 2025 ▶ 4:46 All things AI w @altcap @sama & @satyanadella. A Halloween Special. 🎃🔥BG2 w/ Brad Gerstner · Bg2 Pod

Altman expects coding AI to execute multi-day tasks by 2026

More from Sam Altman

Turn any episode into a week of clips.