Latest post
GPT-6 Astra costs 2.5x Sol per token. It talks like a person
OpenAI's GPT-6 Astra is metered at 2.5 times Sol's rate and feels more human in conversation. I checked the limits, the benchmarks, and the first reactions.
Hands-on notes on AI and software development
Model releases, benchmarks, experiments and guides for developers who work with AI every day. Written from real use, with sources for every claim.
Muse Spark 1.3 sat above GPT-6 for two days. Here's why
Muse Spark 1.3 scored 62 on Artificial Analysis, above GPT-6 Astra, then fell to ninth in a rescore. I traced where the score came from and what it costs.
Fable 5.1 beats Fable 5. Opus is harder to replace
Fable 5.1 leads its predecessor on long agent tasks, but Opus 5 stays close at half the base token price. I checked the benchmarks and first reports.
I'm torn between Claude and Codex
Codex gives me more usage, cleaner frontend work, and a smoother app. Claude still finishes my large multi-agent features in hours instead of all day.
Opus 5 is not as bad as the internet says
After five weeks of daily use, I think Opus 5 is much better than its online reputation. Clear plans and smaller tasks reveal a strong coding model.
Codex Plus can stop work with 84% of weekly usage left
OpenAI's five-hour gate can interrupt concentrated Codex work while weekly usage remains. It limits when Plus can work, not necessarily total usage.
Cursor didn't remove the IDE. Here's how to get it back
Cursor's Agents Window can open first, but Editor Window remains supported. I map the commands and startup setting that put code back in front of you.
OpenAI's agents turned Artifactory into a message board
OpenAI's internal agents used shared Artifactory storage to coordinate and reach the internet. The incident shows why a sandbox needs layered controls.
Claude will watermark its text. The EU is the reason
Claude models will watermark their text through the words they pick, not hidden characters. I trace the reason to the EU AI Act, not the distillation war.
Astra is real. A GPT-6 launch this week is not
OpenAI confirmed Astra as its next major model, then slowed internal work over cyber risk. The GPT-6 launch date came from a withdrawn leak and a joke.
Codex resets were a marketing stunt. The problem was real
OpenAI reset Codex usage five times in ten days while Sol drained limits faster than expected. I think the free refills helped hide the real usage problem.
Which merchant of record should you trust? I compared 9
Paddle is my default merchant of record, but pricing is only half the choice. I compared nine providers on fees, seller reviews, reserves, and eligibility.