Zachi: Cognition's Devin Fusion chart is "a crime"
Cognition says Devin Fusion scores 68 on the Coding Agent Index at 31% lower cost per task than Claude Code with Opus 5.5. Zachi says the chart behind it drops every model under $5, which is most of the field.
Writing criteria
Sydney Runkle's order for fixing an agent: context, harness, then weights
Sydney Runkle wrote up eight takeaways from a conversation with Harrison Chase and Harry Stebbings, built on agent = model + context + harness. Her ordering: optimizing context is the highest ROI move because that is where domain knowledge lives, then the harness, then the weights.
memory is still massively unsolved, but also massively important
The case against
Zachi says Cognition's Devin Fusion chart cuts out the cheap models
Cognition posted that Devin Fusion scored 68 on the Artificial Analysis Coding Agent Index with an Opus 5.5 lead and a GPT-6 Luna sidekick, beating Claude Code with Opus 5.5 at 31% lower cost per task. Zachi says the chart leaves out over 30 models priced below $5, which moves the attractive quadrant and widens the apparent gap.
I like the cognition folks. But this chart is a crime.