Claude open-ended problem success >70%
Claude's success rate on open-ended problems soared from below 20% before December 2025 to well above 70% after the release of Opus 4.5/Mythos.
Snapshot · This Week in Tech (Audio)
TWiT 1087: Evil is the Root of All Money - Could Local AI Laptops Compete With Data Center Giants?
Explore episode
Jun 8, 2026
Where this was said
Anthropic's Self-Improvement Claim: Claude Writes Its Own Code
At 13:20 · chapter starts 11:40
Leo walks through Anthropic's remarkable internal data: engineers are shipping eight times as much code per quarter as in 2021 [1] — Leo Laporte "Claude authors 80%+ of its own code: As of May 2026, more than 80% of code merged into Anthropic's codebase was authored by Claude — up fro…" 12:37 , and as of May 2026, more than 80% of merged code is written by Claude. The success rate on open-ended problems has rocketed from below 20% before December 2025 to over 70% after Mythos launched [1] — Leo Laporte "Claude authors 80%+ of its own code: As of May 2026, more than 80% of code merged into Anthropic's codebase was authored by Claude — up fro…" 12:37 . Fr. Robert raises an important methodological objection — counting commits was famously gamed at Twitter/X under Musk, and the same incentive to inflate metrics could apply here. Jeff Jarvis points out that success rates are measured against moving targets, making the data hard to interpret objectively. The panel reaches a shared suspicion: the timing of this paper, right before an IPO filing, is not coincidental. Leo does not fully dismiss it — the rapid model release cadence from all major labs suggests something real is accelerating.
Claude authors 80%+ of its own code
As of May 2026, more than 80% of code merged into Anthropic's codebase was authored by Claude — up from low single digits before February 2025.
Incentives Over Ideals: Why No AI Pause Will Ever Happen
Fr. Robert Ballecer invokes Charlie Munger's maxim — 'show me your incentives and I'll show you your outcomes' — to explain why calls for an AI pause are performative. The race to build the first fully functional AGI has no second-place prize, making any genuine pause structurally impossible regardless of rhetoric.
Similar snapshots
50 free tools built total
Bhanu and his team built approximately 50 free tools to attract search traffic, each linked back to SiteGPT.
New tool built in under 5 minutes
With AI coding tools like Cursor, Bhanu can now create a new free marketing tool in less than 5 minutes by referencing existing tools.
KD filter < 10 for rankable keywords
Bhanu filters Ahrefs keyword results to show only those with a keyword difficulty below 10, making them realistic ranking targets for any decent website.
Minimum 1,000 monthly search volume
Bhanu sets a minimum search volume of 1,000 monthly searches when selecting keywords to target with free tools.
20 downloads/day at launch
PropGPT averaged 20 downloads per day right after launching on the App Store through influencer marketing.
4 months of rebuilding with no marketing
Eyal and Yali shut down all marketing and spent 4 months completely rebuilding PropGPT from scratch.
Over 40,000 downloads total
PropGPT has accumulated over 40,000 total downloads since launch.
LLM costs only $20/month
PropGPT's large language model (AI) operating costs are just $20 per month, and the cost is continually falling.
Ad monetization for games
Ad-based monetization works well for game apps where users spend extended time in-session, as seen with Grid and Wordle.
