Tuesday, 21 July 2026 Sign in

ShinkaEvolve in Action: How a Human-AI Partnership Conquered a Coding Challenge

Sakana AI ● Covered by 5 sources

Takuya Akiba's competitive programming team used Sakana AI's ShinkaEvolve framework to optimize their code for the 2025 ICFP Programming Contest, winning first place. The system performed 320 trials at a cost of $60, achieving up to 10x speedup on their SAT solver by discovering a more efficient intermediate representation for the maze topology encoding. This collaboration enabled the team to solve previously intractable large-scale problems and provided insights that the human programmers then applied to other challenges.

Read the full story at Sakana AI

Also covered by

Related stories

ShinkaEvolve: Evolving New Algorithms with LLMs, Orders of Magnitude More Efficiently Sakana AI

Introducing Sakana AI’s Recursive Self-Improvement (RSI) Lab Sakana AI

Sakana AI Wins Award at US-Japan Competition for Defense Innovation Sakana AI

Accelerating AI Implementation with AI: Sakana AI Software Engineer Interview Sakana AI

Cutting-edge AI Technology for Business: Sakana AI Applied Team Member Interview Sakana AI

[Sakana AI Applied Case Interview] Behind-the-Scenes Development for AI Agent Implementation in Banking Operations Sakana AI

The daily briefing

Every AI story that matters,
in your inbox by 8am.

TLDRocket reads 60+ sources, removes duplicate coverage, and summarises the day in two minutes. Free, no spam, unsubscribe anytime.

TLDRocket 20 July 2026

The day's AI story splits into two competing visions of the future, each with immediate consequences. Google is betting everything on custom silicon: its new Frozen v2 chip, designed specifically for Gemini, promises six to ten times better efficiency per watt, sending the stock up 3 percent as investors see validation for the company's $180–190 billion AI spending strategy. This follows NVIDIA's release of Cosmos 3 Edge, a 4-billion-parameter world model for robotics and edge devices, signaling that specialized hardware tailored to specific workloads is becoming the competitive advantage. Yet that same calculus is being upended by Chinese competition. Moonshot AI's Kimi K3, a 2.8 trillion-parameter open-weight model released this week, matches Anthropic's Claude Fable 5 line-for-line on coding tasks at one-third the cost—though four times slower. The model ranks among the world's best performers, demonstrating that frontier capability no longer requires closed weights or American dominance. OpenAI executives are pushing for US regulation of open-weight models, fearing the business model erosion Kimi K3 represents. But the underlying tension is structural: as infrastructure commoditizes and Chinese labs execute better than American ones on similar architectures, the advantage shifts from who controls the chips to who controls the standards. The Model Context Protocol's update to stateless session management, buried in today's news, matters more than it appears—it's the unsexy infrastructure that lets any model talk to any service, regardless of who trained it or where it runs.

Read the full briefing →