In the news
Introducing SWE-2: Pushing the Pareto Frontier
Cognition · Published · 3 min read · 342 on Hacker News
In 30 seconds
- What happened
- Cognition released SWE-2, a coding model achieving 50% on FrontierCode 1.1 while costing 64% less than comparable competitors.
- Why it matters
- Engineers building or evaluating AI coding agents need to assess whether SWE-2's cost-performance tradeoff fits their deployment constraints and use cases.
- Watch out
- Benchmark results vary significantly across different coding tasks; SWE-2 trails GPT-6 Astra on some benchmarks and excels on others, so real-world performance depends heavily on your specific problem domain.
The patterns behind this
Each one covers how the technique works, when it earns its cost, and where it breaks.
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.