In March 2024, Lee wrote about a GitClear report that had analyzed 153 million lines of changed code, and said the early numbers on AI-assisted code were worrying.

GitClear kept counting. The newest report covers 623 million changes through the first half of 2026.

So this episode is a check on his own words: what he got right, what he got wrong, and the thing he missed completely. The metric he led with in 2024 turned out to be the least alarming one in the report. The one he underplayed got worse than he expected.

In this episode:

• The churn number that didn't double, and what projections do
• Duplication up 81%, copy/paste from 9% to 16%, cross-file calls down 35%
• Refactoring collapsed: 21% of changed lines in 2022, under 4% now
• Why "does AI write good code?" is the wrong question
• The price change underneath it all: writing got nearly free, understanding didn't
• Two fair objections, and what he'd write today instead of "use it in moderation"

Links:

• Read the original 2024 article
• GitClear: Coding on Copilot, the 2024 report
• GitClear: The Maintainability Gap, the new report
• Get the SAI newsletter
• Atchison Academy courses

GitHub published five incidents for August 2026. Three of them are the same failure.

On the 17th, seven and a half hours of degraded service: 29,000 organizations, 4.8 million requests failed or slow, 56% of front-door requests failing at the peak. Three weeks, three incidents, one blind spot — capacity ran out in places nobody was graphing.

Your autoscaler scales the thing you are observing. Saturation happens in the thing you don't.

In this episode:

• The August 17th outage, and the two beside it on the 6th and the 26th
• Sidecar concurrency, network flow tables, database primary headroom: limits with no dashboard
• Why these limits are dangerous: someone else's layer, hard rather than soft, invisible
• Retry amplification, and how one data center's capacity problem became an authentication outage everywhere
• Protective throttles set so high they engaged after the damage was done
• Two lists to build this week: your real hard limits, and your retry code

Links:

• GitHub Availability Report: August 2026
• Get the SAI newsletter
• Atchison Academy courses

It's 3am and a payment service is failing. The engineer who answers has never touched it, and the next hour goes to finding someone who can fix it.

In this episode, Lee goes further than his article "When Everything Is Critical, Nothing Is." The article named the two missing decisions. This episode is about how to make them: which services matter most, and which one team owns each of them.

In this episode:

• The four service tiers, in plain terms
• Three questions that keep everything from landing in Tier 1
• Why every tier is a budget decision, and why you should cap Tier 1
• The dependency trap: a Tier 1 service is only as reliable as the lower-tier service it can't live without
• What "owner" means: paged, planned, permitted
• Inherited services, and an exercise to run this week

Links:

• Read the original article
• The ownership-gap diagnostic
• STOSA
• Get the SAI newsletter
• Atchison Academy courses