Sentry founder praises Opus 5.5 for engineering, shares agent test fixture design details
zeeg · x · 2026-10-02
Sentry founder David Cramer says he's "really impressed with Opus 5.5," calling it truly great for engineering tasks after what he saw as several weak Anthropic releases, and shared the model's output on getsentry/junior.
The attached GitHub issue details his agent testing methodology: a single small fixture where every turn runs the real agent and real model, using history instead of mocks. Data shows 61% of commits touching agent code in the last six months also changed these tests; there were 8 ways to set up a run, 11 ways to fake the model, and 5+ ways to load history. Fake-model tests still sent 320 live AI Gateway requests per run (1.7–7s each), causing most local timeouts.
Related event: Sentry founder praises Opus 5.5's engineering prowess(2 posts)→
More from coding & agent
- George Hotz seeks a university to host his hands-on ML compilers course; UIUC student says he can make it happen — ctjlewis · 2026-10-02
- Notion ships official agent setup guide positioning it as durable shared memory for AI agents — nbaschez · 2026-10-02
- Non-coder builds full 3D world in 12 hours using 93% of a $20 Claude weekly quota — FinanceYF5 · 2026-10-02
- Coding Agent Index: Sonnet 5.5 tops at 68 but costs $14.19/task vs GPT-6.1 Sol's $1.04 — ArtificialAnlys · 2026-10-02
- Stitch agents now improve this developer's products around the clock — tokumin · 2026-10-02
- Pi 1.0 released with MCP support enabled by default — psychohistorian8 · 2026-10-02