Agent Selection Strategies for Large-Scale Engineering
silasalberti · x · 2026-07-14
Ryan Carson shared his workflow: he completed almost all engineering tasks yesterday using SWE 1.7; he used Fable once for a tricky, time-sensitive issue; and he turns to Fusion when he needs more intelligence than SWE 1.7 but doesn't need the absolute strongest model.
He believes developers still use CC or Codex largely due to token subsidies. However, once a team scales into a "full software factory," it becomes crucial to optimize both cost and intelligence. The best approach is using an independent agent lab or building one in-house like Ramp/Inspect, though the latter is only viable at a certain revenue and team size.
More from coding & agent
- A better path to agent autonomy is running waves, finding friction, and iterating — JnBrymn · 2026-07-22
- Coding agents are heading toward an AI-writes, AI-reviews, human-approves workflow — aftahi_ai · 2026-07-22
- oMLX 0.5.2 adds Mac menu-bar stats, low-bit decode kernels, and faster downloads — awnihannun · 2026-07-22
- GitHub review bot hits its PR limit and forces a 39-minute cooldown — DanielLockyer · 2026-07-22
- Max reasoning effort appears to be mobile-only in Codex Remote, not desktop — GabGarrett · 2026-07-22
- A Reddit demo argues online stores should expose carts and pricing through MCP — gelembjuk · 2026-07-22