Posts

Engineering ·
Reducing LLM Calls and Fixing the Cache and Morning Batch
With 1.6 days of credits left, how regeneration, reasoning tokens and call counts were cut, and how a cache that behaved differently from what we intended got fixed.
Engineering ·
Finding What Vector Search Missed with Keyword Search
When the documents grew 4.5 times, vector search alone dropped to 0.375 Recall@5. This post covers Korean bigram keyword search, RRF, and title, source and freshness weights, alongside stage-by-stage results and their comparison limits.
Engineering ·
When Output Checks Threw Away Good AI Answers
FINCH runs 10 automatic checks before sending an LLM explanation. Counting rejections showed many were false positives, mostly caused by the prompt and the checks disagreeing, not by the model.
Engineering ·
Why Janus Stopped Giving Agents a Separate Worktree
Janus first gave every Task its own git worktree. To see results right away it now works in the original checkout. How the risks git can't undo were documented and contained.
Engineering ·
Why Janus Runs Three Concurrent Generations, Not Five
On a 48GB Mac running a 27B model, concurrent generation requests went from 5 to 3. The reasoning on KV cache, a single GPU and MTP, and the measured condition for raising it again.
Engineering ·
Runtime Contracts and Infrastructure Ownership Boundaries for Deployment
Runtime contracts, infrastructure ownership, and the gap between passing checks and deploying: a look back at PinLog’s July 2026 records.
Engineering ·
PVC Retention Policies and Data Recovery Limits on Single-Node k3s
Looking at local-path and local-path-retain on a single-node k3s cluster, I separated three questions: Pod restarts, PVC deletion, and recovery after losing the node.
Engineering ·
An External Monitoring Path That Works Even When the Node Fails
Internal alerts cannot watch over the complete failure of the node they run on. How PinLog separates internal monitoring from external HTTPS/TLS checks, and what that separation does not solve.
Engineering ·
A CI Refactor of Waiting PRs, Duplicated Config, and Coverage Gates
Starting from a Channel.io blog post: auto-merge, composite actions, coverage gates, and pre-checks before building, across five sprints. How one reference article turned into a real pipeline overhaul, and why two pieces of work that shipped zero lines of code were still wins.
Engineering ·
Unifying JWT, Cookie, and Frontend Login Expiry Policies
The JWT lifetime is 2 hours, but users get logged out after 1. Cookie lifetime, sliding refresh while the user is active, a frontend timer: 4 layers each had their own clock. Here's how we unified them into a single source of truth: server env vars in one place.
Engineering ·
A 15-Job CI Pipeline for Verifying AI-Written Code
The faster AI wrote code, the tighter the verification had to be. A hands-on account of building 15 CI jobs, from secret scanning to GitOps locks.
Engineering ·
MSA Service Boundaries Designed for Splitting Work Among AI Agents
A hands-on record of MSA design where humans make architectural decisions and AI carries them out