Claims

Prediction · AI capabilities

“for coding, we're going to see very serious things by the end of 2025,”

Dario Amodei, Anthropic CEO · 28 February 2025 · Hard Fork - Anthropic's C.E.O. Dario Amodei on Surviving the A.I. Endgame Transcript (Podscripts (Hard Fork))

By the end of 2025, AI will do very serious things in coding.

Resolved true The record shows it happened or was so.

What would count as true or false

'Very serious things' in coding means AI going well beyond autocomplete to substantial, largely autonomous software engineering. True: by 31 Dec 2025, evidence shows AI systems doing substantial real-world coding work. Examples: widely deployed agentic coding tools that complete multi-file or multi-hour engineering tasks with little supervision, credible reports that major companies get a large share of their code from AI, or top scores on demanding coding benchmarks such as SWE-bench Verified at roughly 70% or above. Mostly true: clear major advances that fall somewhat short of autonomous serious engineering, for example strong benchmark results but limited real deployment. Mixed: some advances but mostly incremental. Mostly false: AI coding stayed mainly assistive (suggestions, small snippets), with only demos of anything more. False: dated evidence from late 2025 shows AI coding capability and use essentially unchanged from early 2025.

Evidence

1. Anthropic acquires Bun as Claude Code hits $1B Anthropic · 3 December 2025 Own record

   “In November, Claude Code achieved a significant milestone: just six months after becoming available to the public, it reached $1 billion in run-rate revenue.”

2. Microsoft CEO says up to 30% of the company's code was written by AI | TechCrunch TechCrunch · 29 April 2025

   “During a fireside chat with Meta CEO Mark Zuckerberg at Meta’s LlamaCon conference on Tuesday, Microsoft CEO Satya Nadella said that 20% to 30% of code inside the company’s repositories was “written by software” — meaning AI.”

3. SWE-Bench Verified WorldNeural

   “| Claude Sonnet 4.5 → | 77.2 | Self-reported * | 2025-09-29 | View source ↗”

4. Is 90% of code at Anthropic being written by AIs? — LessWrong LessWrong (post by ryan_greenblatt) · 22 October 2025

   “This is a reasonable summary of what Dario said when talking to Marc Benioff, but my current understanding is that it isn't true that AIs are writing 90% of code at Anthropic.”

Reasoning

The criteria's examples of "very serious things" in coding are met by 31 Dec 2025. In April 2025, Microsoft's CEO said 20–30% of the code in the company's repositories was written by AI (E2). A third-party aggregator lists a 77.2% SWE-bench Verified score for Claude Sonnet 4.5 dated September 2025, though the score is self-reported (E3). Anthropic's official announcement says Claude Code reached $1B in run-rate revenue by November 2025, which points to wide real-world use of an agentic coding tool (E1). E4 questions the specific claim that 90% of Anthropic's code is AI-written, but it does not undercut the broader capability and adoption shown here.

Cite this claim

Dario Amodei, “By the end of 2025, AI will do very serious things in coding.” (said 28 February 2025). Verdict: resolved true. Reliability API, claim c_ec64e8bb2ea9, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_ec64e8bb2ea9