Geoff Krombach

"Capitalism" sounds evil Implying the system is for the benefit of money and those who have it I think we need a new -ism word for the system that (at its best) enables vast networks of cooperation, and everyone getting more of what they want

Perplexity Computer can now help prepare your federal tax return. Select “Navigate my taxes” on Computer to give it a shot.

Bought a new iPhone (17) - about 3 years since I last upgraded. I thought the process of setting up a new phone from an existing one would be massively improved in that time, but no, it's still a slow, clunky, over-complicated mess. 2 hours in and I'm still faffing.

Codex update 0.106.0 is out! This update gives us a shiny new toy. The "Ask Question" tool enabled in default coding modes! This enables custom plan modes, skills, and "interview me" style brainstorming sessions not possible without this feature. The great news is, if you don't want to use it, it's off by default. To enable it add the feature flag: default_mode_request_user_input = true Agent Memory Updates 105 and 106 are also PACKED with agent memory improvements. Memory now does diff-based forgetting so stale thread memory gets surgically pruned Memory selection is now usage-aware so active high-signal memories are prioritized. I have found memory to be working very well now. Javascript REPL Other update to speak of, Javascript REPL is now under the /experimental menu, allowing for interactive website debugging. Voice Transcription Still no voice transcription for Linux. I have not tested Windows. Please let me know in the comments if you try it. v2 Websockets Fixed responses_websockets_v2 now fully works on its own, including websocket prewarm and request routing. No more error/mismatch behavior. responses_websockets_v2 = true Very well done from the Codex team. Thank you!

Just woke up and checked it! Overnight: → 3 major bugs fixed → all benchmarks improved (from 4.8% to 92.8% !!) → total code size *reduced* by 10% I just set up a GOAL.txt with optimizable metrics and left a crowd of AI's improving the Ben2 repo. Very cool

I've been using OpenClaw exclusively as my only AI tool for the last three weeks. For content, I have an agent on Gemini 3.1 Pro for everything except social media posts where I have a dedicated agent on Opus 4.6. But today I hit a wall. For the last week I've been working on a project with my Gemini 3.1 Pro agent that required referencing several markdown files with ±100,000 words of source material that I either wrote or spoke and transcribed with speech-to-text. I've been sparring with the Gemini agent, going back and forth, refining sections, and frankly making painfully incremental progress. Then on a whim, I fed the source material into a project in the Claude Chat app on macOS and picked Opus 4.6 Extended to see if it could do a better job. My jaw hit the floor. It one-shot an absolute masterpiece. Opus 4.6 Extended took my voice and ideas from the source material, structured and polished it in a way that would have taken me 15 hours of back-and-forth with my Gemini agent on OpenClaw. The output was 97% of the way to a final draft, and it only took one messy spoken prompt plus the source material to get there. I don't know what wizardry Anthropic are doing, but it's working. (There's also a chance I was experiencing context rot with this task on Gemini 3.1 Pro through OpenClaw.) Regardless, I had to pick my jaw up off the floor. My use of Opus 4.6 through OpenClaw hasn't been as impressive as this experience with Opus 4.6 Extended in Claude Chat. I wonder if Anthropic is keeping some sort of special firepower for its own apps. If that's the case, I'm tempted to move to a full Claude Code setup, because the difference in this use case was that significant. Having the ability to think through and use such insane volumes of source material opens up a world where you can synthesise and condense unfathomable amounts of information into finished content in a few seconds. We're living through the golden era.