Journal.

The most expensive mistake: $2922 in one session. And it's not her fault

Back in June I wrote about watching Claude and Codex ping-pong while paying for tokens. Promised to tally it up over time. Did. Not by choice: in two days the limits on both accounts hit zero, so I asked the main agent to find out where it all went. Sixteen…

Memory chaos in agents. Part 2: I fixed it three times

Part 1 was about what the audit revealed. Now the uncomfortable part: some of the breakage I had already fixed before. I use Codex as a second opinion: Claude writes the plan, Codex reviews it. And that Codex was dumping notes into the archive of one specific project across all projects…

Memory chaos in agents. Part 1: 2204 commands and zero entries

I asked the bug-hunter agent how many entries it had in memory. It said: zero. A month and a half of work. 2204 terminal commands. Labels, donor sync, a spec with 55 fixes. And zero entries. It wasn't being lazy. The folder for its memory didn't exist…

Going? Then I'm pulling out the swap model and assembling a diverse dataset, showing

a contact sheet for approval before training. Baked for 1m 20s — did you choose this idea just because it's easier? Honestly — yes, partly. I leaned on the fact that insightface was already installed, and that was laziness, not the best choice. Thanks for pointing it out.…

How long did it take to find your angle, dream of hanging in the net, focus on something meaningful……

Hreflang × subdomain × subdirectory: what to pick

A four-language site can be built three ways: ccTLD (example.de), subdomain (de.example.com), or subdirectory (example.com/de/). Each has its own upsides and pitfalls. What I picked and why — after a month of wrestling with hreflang.

JSON-LD Schema.org: Which Types to Choose

Schema.org tells Google what's on your page. JSON-LD is the format Google prefers. BlogPosting / FAQPage / HowTo / Product / Organization / BreadcrumbList — which to pick for each page type. Based on marking up ten projects.

robots.txt: what to block, what to allow

A simple text file that breaks SEO with one wrong rule. Disallow / Allow / Sitemap / User-agent. How robots.txt differs from noindex and where confusing the two is dangerous. From hands-on experience across ten production sites.

Canonical URL: When You Need It, When It Hurts

Canonical tells Google: 'this is the main version of the page.' Useful for duplicates, filters, A/B tests. But if you set it up wrong, Google stops ranking the original. Here's what I found while debugging canonical tags across five projects.

H1: why one and how long

Google says multiple H1s are fine. But in practice, one works better: cleaner semantics, easier hierarchy, better accessibility. Here is what I found while cleaning up H1s across ten projects.