Musings on design, product, psychology, society and business. Often discussing good leadership, (remote) organizations and psychology. With a side of randomness.
Currently: Sr. Director, Product Design EMEA at Xero
Previously: VP Product & Design, Jetpack at Automattic
| Web | https://intenseminimalism.com |
| Newsletter | http://tinyletter.com/folletto |
| GitHub | https://github.com/folletto |
| Pronouns | she/her 🏳️🌈🏳️⚧️ |
Words are cheap. Use fewer.
The arrival of “AI” has created disruption. I don’t mean the one generally in hype, but a subtler one: all the artefacts that were synonymous of intellectual work now have lost that meaning. Essays, reports, analysis, papers, articles: all of them can now be parroted out by an LLM with ease, in seconds. Words are now cheap. A one line prompt can make a whole essay.
https://intenseminimalism.com/2026/words-are-cheap-use-fewer/
Want to have a quick glance about reading time estimate for any web page you are visiting?
I just updated my tiny minimalist Chrome Extension, Blipread.
https://chromewebstore.google.com/detail/blipread-page-reading-tim/pimmlmgpidiceckodhdifoppcidlndbp
New:
• Text selection support
• Popup with estimate total & selection time
• Better error handling
I did some pre-release testing of the new Bear CLI, and now that's it's released I can share.
Here's a Bear skill and MCP, based on the official CLI: https://github.com/folletto/bear-llm-toolbox
Honestly I was surprised to see how well it worked.
Disclaimer: this is likely needing some polish.
Leaking Claude Code harness means that the open source harnesses will now become more similar to Claude Code and better suited for Claude LLMs, giving Anthropic an advantage.
Strategic or accidental leak doesn’t matter anymore.
Almost doubled the speed in running locally on Apple Silicon.
Anyone wants to put bets on when local models will take over?
13 years ago I made a little MacOS app to launch different Firefox profile easily in Objective-C.
11y ago I rebuilt it from scratch in Swift to practice.
Today I rebuilt (and much more) with Claude Code for macOS. I used again to learn: SwiftUI.
If you use Firefox profiles, you might like it.
Running LLM models locally on common hardware is getting closer...
"TurboQuant proved it can quantize the key-value cache to just 3 bits without requiring training or fine-tuning and causing any compromise in model accuracy, all while achieving a faster runtime than the original LLMs"
https://research.google/blog/turboquant-redefining-ai-efficiency-with-extreme-compression/
RE: https://mastodon.social/@folletto/113356958558592033
Still unfixed. Sigh.
Chrome: trust me, update is safe, I will reopen your windows correctly in your many virtual desktops
Also Chrome: opens all the windows in the same desktop.
FFS.
What does it mean to approach AI (or better, LLMs) as a material in our product and design work?
Here's an article to help building with LLMs, in whichever product area you work.
https://intenseminimalism.com/2024/a-quick-primer-on-designing-with-ai/
