Weekly read of high-signal AI builder discourse on X. Not a news summary — a vibe check. What's the room feeling, what's getting retweeted, what's being quietly walked back.
What's loud this week
Coding assistants are still the dominant topic by volume but the conversation has shifted. The "is it actually useful" debate has largely resolved in favor of yes — the new fight is about which one and for what task. Codex vs Claude Code vs OpenCode is a weekly recurring event at this point, with increasingly granular task-specific benchmarks replacing the broad "which is smarter" takes that dominated Q4 2025.
Agent frameworks are having a genuine moment. Not the hype cycle from 18 months ago — more grounded, more production-focused. People are shipping agents and talking about what broke in the first week of real use rather than what the demo looked like.
What's getting quieter
Long-context hype is deflating. The dream that 1M token windows would solve retrieval has run into the reality that most models degrade past 100K in useful ways, and that structured retrieval is still cheaper and more reliable for most production workloads. The RAG-is-dead crowd has largely gone quiet.
General "AGI is coming" posts are also down in engagement. Not gone — but the audience has moved on to more specific conversations. Builder Twitter has less patience for it than it did six months ago.
Signal worth watching
Education AI is a low-volume topic that's showing consistent week-over-week growth in quality of discourse. Not teachers complaining — builders shipping. Tutoring systems, personalized curriculum, adaptive assessment. The market is moving but hasn't shown up in mainstream AI coverage yet. That gap usually closes fast.
Next signal check publishes Friday. If you want to flag something worth including, the repo is open.