
Amodei said AI would write 90% of code within months.
Dario Amodei · CEO · AnthropicSee track record
“I think we'll be there in three to six months, where AI is writing 90 percent of the code. And then in twelve months, we may be in a world where AI is writing essentially all of the code.”
- Confidence
- 56/100
Disputed
- For vs against
- 1to0
Nothing contradicts it yet
- Days tracked
- 471days
Since first seen on the record
- Saved
- 1,460
Readers tracking this signal
Will this come true?
48% say true, 16% too early, to 487 votes
A claim about the future
Within three to six months AI will write 90% of all code, and within a year essentially all of it.
By the six month and one year marks, AI was not authoring 90% of production code in the simple way the line implied. Studies showed mixed productivity and more security flaws. The raw line-count framing was arguably defensible, but the blanket claim was too facile.
See what happenedIt is the most cited timeline for AI replacing engineering work, and a benchmark for how aggressive lab leaders' near-term forecasts run.
How the claim has aged.
- Where it beganMar 10, 2025
Statement made
Council on Foreign Relations event
- VerdictJun 20, 2026
Partly right
By the six month and one year marks, AI was not authoring 90% of production code in the simple way the line implied. Studies showed mixed productivity and more security flaws. The raw line-count framing was arguably defensible, but the blanket claim was too facile.
- Supports
- Contradicts
- Neutral
Show the receipts.
- QuotedSupportsMar 10, 2025
Dario Amodei · Anthropic · Interview
I think we'll be there in three to six months, where AI is writing 90 percent of the code. And then in twelve months, we may be in a world where AI is writing essentially all of the code.
- PatternContext3d ago
What happened since · Filing
By the six month and one year marks, AI was not authoring 90% of production code in the simple way the line implied. Studies showed mixed productivity and more security flaws. The raw line-count framing was arguably defensible, but the blanket claim was too facile.
Right on direction, loose on timing. The headline number was a framing choice, not a measured fact.
It is the most cited timeline for AI replacing engineering work, and a benchmark for how aggressive lab leaders' near-term forecasts run.
- Independent data shows AI authoring most merged production code.
- Engineering headcount falls while output holds.
- Human review and new code creation keep people central.
- Reliability caps how much ships unedited.
The next proof point.
How it resolves
Whether the mixed record tips one way.
Next frontier release
Whether capability or integration drives the next leap.
Agent deployment data
Whether agents move from pilots to measurable output.
Commit a view. The record will remember whether you were early, or wrong. Both are worth knowing.
Call it, and argue it.
No calls yet. Be the first.
Keep the receipts on this one. Save it, track the topic, and follow the arcs it belongs to. Then come back when the next proof point lands.
- The AGI TimelineContested
Follow the thread.

Agents “join the workforce”
Sam Altman · Jan 6, 2025

“Smarter than any human” by 2025
Elon Musk · Apr 8, 2024

The delayed Siri overhaul
Tim Cook · Oct 30, 2025