Sumitsuke

Design, verify, complete. Auditing and repairing AI-generated / outsourced code; measurements with frozen logs and independent recomputation. Lab: sumitsuke.jp/lab — GitHub: sumitsuke

Joined on Jul 7, 2026 https://sumitsuke.jp/

Sumitsuke

Sep 21

I cut transfer by 81–257 KiB. Did LCP actually improve? 30 Lighthouse runs

#performance #lighthouse #webperf #testing

4 min read

Want to connect with Sumitsuke?

Create an account to connect with Sumitsuke. You can also sign in below to proceed if you already have an account.

Create Account

Already have an account? Sign in

Sumitsuke

Sep 21

Can an LLM measure UI thresholds from screenshots? 164 crops vs DOM gold labels

#llm #a11y #css #testing

5 min read

Sumitsuke

Sep 21

I asked an LLM to cut 14 texts by 20–30%. Median: only 7%

#llm #prompt #testing #writing

5 min read

Sumitsuke

Sep 17

I let an AI write a technical article with zero human editing, then ran it through my normal verification pass

#ai #writing #testing #review

7 min read

Sumitsuke

Sep 16

A JSONL record split in two: U+2028, U+0085, and the separator I missed

#python #json #debugging #testing

4 min read

Sumitsuke

Sep 15

Does AI-generated code silently swallow errors? 120 measured generations: every flagged case was a false positive or a documented fallback

#ai #llm #staticanalysis #python

21 min read

Sumitsuke

Jul 12

LLM self-review vs. a blind re-run on 160 gold-labeled name-matching pairs — no measurable difference at ceiling

#ai #llm #testing #nlp

7 min read

Sumitsuke

Jul 9

What actually breaks when a free Cloudflare D1 database hits its 500 MB wall

#cloudflare #database #sqlite #serverless

6 min read

Sumitsuke

Jul 7

I had a local LLM review the same code 51 times: majority voting picks persistence, not correctness

#ai #llm #codereview #testing

13 min read