AI Private Deployment · DevOps in Practice · The One-Person Company
This is Alex's technical blog: hands-on notes on private LLM deployment, DevOps and cloud-native engineering, and the real experience of building products with AI as a one-person company.
Latest posts

I Found TypeSafe on HN After Two Years of Silence—Don't Chat. Decide.

It Was 9 PM Friday, I Was Done, and the Dashboard Had Other Plans

Three Days Staring at server_is_overloaded: What I Should Have Tried First

The Afternoon I Saw the Same ChatGPT Line Three Times and Finally Understood It

The Night I Read Dario Amodei's Case for Pacing Frontier AI
Popular posts

After two years of radio silence, the RLHF co-inventor behind ChatGPT posted four words on X and the HN upvotes started climbing.

I was five minutes from calling it a night when the monitoring dashboard showed me something I wasn't expecting.

Three days of the same red error, the same panic loop, until one small config change finally broke the cycle.

Three times in one afternoon, the same line hit my ChatGPT screen, and by the third one, I finally stopped scrolling.

A late-night read turned into a long conversation about whether we're moving too fast with frontier AI.

After quoting an OpenAI internal note as if it were gospel, I sat with the discomfort of having gone a bit too far.
All posts292
View the full archive
After two years of radio silence, the RLHF co-inventor behind ChatGPT posted four words on X and the HN upvotes started climbing.

I was five minutes from calling it a night when the monitoring dashboard showed me something I wasn't expecting.

Three days of the same red error, the same panic loop, until one small config change finally broke the cycle.

Three times in one afternoon, the same line hit my ChatGPT screen, and by the third one, I finally stopped scrolling.

A late-night read turned into a long conversation about whether we're moving too fast with frontier AI.

After quoting an OpenAI internal note as if it were gospel, I sat with the discomfort of having gone a bit too far.

One morning my Google account just wasn't there anymore, and the ripple effects hit everything I thought was safe in the cloud.

I didn't plan to read it, but one OpenAI insider's tweet turned my lazy afternoon scroll into something I couldn't shake.

The notification hit at 3:15 a.m., my hands moved before my brain caught up, and by 4:00 the whole stack was on Astra.

GPT-6 Astra is officially live, and after digging in, here's what genuinely surprised me about the new model.