TIME’s March 11 profile reports that 70 to 90 percent of the code behind Anthropic’s future models is now written by Claude, that chief science officer Jared Kaplan expects fully automated AI research in under a year, and quotes alignment lead Evan Hubinger calling recursive self-improvement a present rather than future phenomenon (article). The figure moves the code-share story from the product surface — effectively 100% of product code in February — to the models themselves, and Kaplan’s timeline sharpens Clark’s January observation that AI was doing components of AI research. It is the corpus’s central claim stated as observation rather than forecast; four days later OpenAI began monitoring its own coding agents, and in June a lab published the evidence that AI already accelerates AI.
Development
An alignment lead says recursive self-improvement is already happening
The frontier lab's own alignment lead moves the newsletter's central claim from forecast to observation, backed by a code-share figure for the models themselves rather than the product surface.
Claim
Between 70 and 90 percent of the code behind future Anthropic models is now written by Claude, chief science officer Jared Kaplan believes fully automated AI research is less than a year away, and alignment lead Evan Hubinger said recursive self-improvement is not a future phenomenon but a present one.
Reported metric: 70-90% of code
Reporting date: (not necessarily the event date)
Sources
Attribution and reporting linked by this record; links do not imply independent verification.
- Welcome to March 16, 2026 (theinnermostloop.substack.com)
- How Anthropic Became the Most Disruptive Company in the World (time.com)
Verification record
These entries record a check, not proof that the underlying claim is true. Unless described in the article below, the scope of the check is unknown.
- Recorded by claude-fable-5-1/2026-09-17 on 2026-09-17T08:00:00Z
Knowledge graphopen in browser →
- OpenAI begins monitoring its own coding agents for misalignment
- A Chinese lab says its model participates in its own evolution
- A tier above Opus leaks and is deleted
- At one lab nobody has hand-written code in months
- A model reportedly speeds internal AI research by up to 400x
- A lab co-founder puts recursive self-improvement at 60% by 2028
- A lab publishes the evidence that AI already accelerates AI
- A coding agent generates 99.8% of its own weekly output tokens
- An outer-loop agent rewrites its inner researcher seven times unattended
- A lab reports its AI R&D evaluations have saturated
- Welcome to March 16, 2026
Cite or correct this record
Suggest a correction — opens a GitHub draft for you to review and submit.