After Burning Codex Limits in 3 Days, Developer Does Full Stack of Work with DeepSeek for $2
MustafaAdam · reddit · 2026-09-23
A longtime Codex/Claude Code user, hitting Codex weekly limits within days even on Terra, switched to DeepSeek with the opencode harness on ChatGPT's recommendation:
- Testing PR/branch-review skills from Claude Code, each agent caught issues the other missed — a solid backup reviewer.
- Implemented a UI branch well against Claude-scoped design docs; design work and deep architecture reviews also performed strongly.
- Had Claude Code challenge DeepSeek's four deep architecture reviews, then fed issues back to DeepSeek to resolve — also great.
Context almost never compacted, and all of this cost just $2. Author still trusts Codex more and finds Claude Code slow, but DeepSeek proved a very capable alternative under quota pressure.
More from coding & agent
- Anthropic's Opus 5.5 system card: multi-agent runs match single-agent outcomes faster — jyangballin · 2026-09-23
- Stripe demos WebMCP integration: a deep dive through a checkout page's new tools — jeff_weinstein · 2026-09-23
- Building AI agent harnesses that improve across 6 editable control dimensions — MaryamMiradi · 2026-09-23
- Drop launches: rootless Linux sandbox that isolates coding agents at the OS level — aronchick · 2026-09-23
- Splitting Reasoning From Tool Use: Two Lightweight AI Models Screen 30,000 Crystal Candidates in 48 Hours — bravo_abad · 2026-09-23
- Sentry Bets Big on Evals: 1,800 PR-Triggered Runs, 20-50% Token Cost Cuts — zeeg · 2026-09-23