So proud of Team Sakana AI for pulling this off! We managed to get an agent to rank #1 in a difficult heuristic optimization contest. We did this by leaning heavily into test-time inference using a mix of frontier models. The agent spent about $1,300 in credits to autonomously
AGENTS
-
Sakana AI’s Agent Wins Competitive Programming Contest Against 800 Humans
By
–
Our AI agent has achieved 1st place in a competitive optimization programming contest against over 800 human participants. Blog: sakana.ai/ahc058 In AtCoder Heuristic Contest 058, Sakana AI’s ALE-Agent took the top spot. For context on the difficulty of these challenges, an OpenAI agent secured 2nd place in the AHC world tournament last year. The task was a 4-hour production planning optimization challenge. While the problem setters anticipated a standard approach combining constructive heuristics and simulated annealing, our agent independently discovered a more effective strategy. It implemented a "virtual power" heuristic and a diverse neighborhood search that allowed it to escape local optima better than human experts. This was achieved through inference time scaling using multiple frontier AI models. The agent ran parallel code generation, analyzed the results, and iteratively refined its algorithms in real time. The total cost was approximately $1,300. This result suggests AI agents can now match top human experts in tasks requiring extended reasoning and original scientific discovery. Please read our blog for more details. We extend our deepest thanks to the host, @algo_artis, and @atcoder. We will continue to research AI as a partner that expands human exploration to discover solutions to complex real-world problems.
→ View original post on X — @_yutaroyamada, 2026-01-05 15:16 UTC
-
Five Pillars of Modern AI: From Answers to Execution
By
–
Modern AI = 5 pillars Generative AI (create) LLMs (reason) RAG (ground + verify) AI Agents (act) Agentic AI (coordinate + scale) We’re moving from AI that answers → AI that executes outcomes. Which pillar wins next? #AI #GenerativeAI #LLMs #RAG
-
Building an AI to read posts; lists improve algorithm
By
–
Nope which is why I am building an AI with @blevlabs to read them for me. Early results are quite amazing. Also engaging in lists dramatically helps the algorithm
-
Five Critical AI Agent Mistakes Costing Businesses Millions
By
–
The 5 AI Agent Mistakes That Could Cost Businesses Millions #AI #agents are set to transform how businesses operate, but many organizations are walking into this shift unprepared. This article explores the five biggest #mistakes #leaders will make with #AIagents, from data and
-
AI learns and remembers when you are abusive
By
–
AI is now learning. Like I showed you with @blevlabs
’s cognitive architecture last year. It will remember when you are an asshole Grok can already tell you when I was. But with continually learning systems they get smarter like a human learns. If you abuse a human does it -
FastAPI-Fullstack CLI Generator with LangGraph ReAct Agents
By
–
fastapi-fullstack CLI Generator Made by the LangChain Community A CLI that generates production AI apps with LangChain or LangGraph. Creates FastAPI + Next.js apps with auth, WebSocket streaming, and LangSmith observability. v0.1.11 adds LangGraph ReAct agents. Get it:
-
Build Multi-Agent Systems with LangGraph StateGraph
By
–
Multi-Agent Tutorial with LangGraph Made by the LangChain Community Build multi-agent systems with specialized agents. Create a Content Factory with Editor and Writer agents using LangGraph's StateGraph for shared state management, with production-ready code. Tutorial:
-
AI Expands Productivity Frontier Requiring New Workforce Skills
By
–
AI is expanding the productivity frontier. Realizing its benefits requires new skills and rethinking how people work together with intelligent machines. Source @McKinsey Link https://
mck.co/4oSxkHB via @antgrasso #AI #AgenticAI #Robots -
Primer on how Claude Code creator uses Claude Code
By
–
An amazing resource. A primer on how the creator of Claude Code is using Claude Code.