All Articles
A Fail-Closed Model Pricing Registry So New Models Don't Quietly Break Your Cost Math
When Opus 4.8 and Haiku 4.5 landed in the Messages API, rates scattered across my code silently skewed the cost rollup. Here is how to centralize per-model rates and fail closed on unknown models, with complete working code.
The Same 429 Wears a Different Face on Each Route: Running Claude Safely over Anthropic Direct and Azure Foundry
With Claude now generally available on Microsoft Foundry, a two-route setup is realistic even for solo developers. Here is how to fold the route-by-route differences in 429s and retry-after into one normalized error type and a single backoff policy.
Trusting a Three-Day-Old Mirror: Stopping Unattended Tasks from Acting on a Stale Working Copy
A persistent clone reused for speed quietly drifts from the remote. Read 'has this article been fixed yet?' from an out-of-sync tree and your unattended task duplicates work or 'succeeds' against an old world. Here is a HEAD-match plus writability preflight, with a self-healing re-clone, in bash.
Fixing Blurry Wallpapers on New iPhones with Claude Code: Safely Growing a Per-Device Resolution Map
Why wallpapers go blurry or letterboxed on brand-new iPhones, and how to collapse scattered device branches into a single source of truth you can extend in one line — walked through as a real Claude Code refactor with before/after code.
When a Tool Result Is Too Big and Melts Your Context Window: Designing Cursor-Based Pagination
When a list tool returns hundreds of rows at once, an agent's context can collapse in a single call. Here is a cursor-based pagination design that keeps tool output small and protects your token budget, with working code.
Stop rebuilding intermediate files every request: reuse the Code Execution container to carry pipeline state
How to reuse the Code Execution container across requests by passing its container ID, so generated files and intermediate results carry over to the next step. Includes the execution-time billing trap and how to handle container_expired safely, with working code.
Let Claude Actually See the Images Your Tools Return — Use Image Blocks in tool_result and Cut Tokens by Roughly 10x
Stuffing a base64 string into a tool_result makes the same image cost roughly 10–20x more tokens. Here is how to return it as an image content block instead, with SDK code, a token-cost estimate, and the gotchas I hit in production.
Failing Loud on Stale Inputs: A Freshness Contract for Unattended Pipelines
How to stop a scheduled, unattended pipeline from silently shipping degraded work when its upstream data is empty or stale. We implement a freshness contract in bash that asserts recency, non-emptiness, and provenance, plus two real pitfalls I hit running Cowork scheduled tasks.
When Context Editing Made My Agent Re-run the Same Search — Field Notes on Clear Boundaries and Cache Invalidation
After turning on Context Editing to auto-clear tool results, the agent forgot what it had just read, re-ran the same tool, and the cache rebuilt every turn so costs went up. Field notes on instrumenting the silent regression and setting trigger, keep, and clear_at_least from measured data.
Three Ways to Start with Claude — Browser, Desktop, and Mobile, and How to Use Each
A clear walkthrough for getting started with Claude on the browser, desktop app, and iPhone/Android — plus how to combine all three so your day flows smoothly. Written from the perspective of a solo app developer who lives across all three.
Borrowing the Trusted Devices Idea for a Solo Setup: Pinning Automated Runs to Devices You Approved
The Trusted Devices feature that landed in Claude Code on June 28, 2026 is for Team and Enterprise. You can't use the feature on a solo plan, but you can borrow the idea. Here is how to identify a device in a way that doesn't break, and stop any automated run that starts from a device you never approved — with working code and the traps to avoid.
Did That Post Actually Go Through? Safely Retrying an Interrupted MCP Write Without Double-Executing
When an MCP write tool call is interrupted by a dropped connection, you can't tell whether the server ran it. Here's why naive retries cause double-execution, and a working wrapper that uses idempotency keys and a reconcile read to retry safely — with examples from an unattended pipeline.