Weeknotes for week 13, 2026: On mysteries, architecture and german beer.
alexwilson.tech/cont...
#weeknotes
Posts by Alex (アレックス)
Weeknotes for week 11, 2026: On Nandos, Costco and unlimited garlic shrimp.
alexwilson.tech/cont...
#weeknotes
I got this email earlier and I thought it was a really nice gesture, and I suspect Microsoft might be starting a trend.
As a person actively trying to be more vulnerable (from near zero) for the past year, I’m struggling to see the upside: I get light mockery and reductionism, and while I’m delighted people are getting *something* from my experience, I’d really love some help.
Weeknotes for week 10, 2026: On manifesting paranoid london, watching czechia challenge korea, and kabukicho's best curry
alexwilson.tech/cont...
#weeknotes
Weeknotes for week 8, 2026: On community, specifications and Voyager.
alexwilson.tech/cont...
#weeknotes
Slack's new activity manager is almost perfect, but has missed the most useful feature: The ability to add *any* channel I'm in into custom views.
Like most Slack users, I turn off channel notifications to avoid constant beeping, but I still want to follow and group important channels!
walked all 40km of the yamanote line today. pretty fun but now everything hurts!
if only your employer had [for some reason] vibe-coded a nextjs-compatible API that runs on Vite, but there's no way that a serious, grown-up company like yours would have done something like that......
You're finally removing the // @ts-nocheck from each typescript file?
wait, does this mean that currently bank robbers have to bring their own $$$ sacks?
please tell me you actually will implement this to vidgrid?
this is wonderful
Weeknotes for week 7, 2026: On Trifon Zarezan, drinking 13.7 drinks and a 3 year TCO forecast.
alexwilson.tech/cont...
#weeknotes
now THIS is podracing!
Yes: Largely consumer sentiment as tracked by YouGov/etc. doesn't currently reflect investor/shareholder aspirations.
However in some sectors chatbots/agents _are_ upending incumbents.
These aren't mutually exclusive statements: Unnecessary friction in products is ever more important to consider.
In many ways, the rise of AI agents and chat-based UX is like a consumer rebellion against poor UX and complicated billing.
Weeknotes for week 6, 2026: On Russian-food, Anti-Action artwork and C4.
alexwilson.tech/cont...
#weeknotes
If I think about all of the things in my life which cumulatively bring me joy, the loudest is the Roland TB 303 which objectively makes everything better.
Weeknotes for week 5, 2026: On two different jurisdictions of tax, and two different types of soup.
alexwilson.tech/cont...
#weeknotes
Weeknotes for week 4, 2026: On a pivot to MCR, and failing to eat Ekiben.
alexwilson.tech/cont...
#weeknotes
Weeknotes for week 3, 2026: In which I procure a doll, stay indoors a bit, and then go to see a superstar DJ.
alexwilson.tech/cont...
#weeknotes
Weekend thoughts on Gas Town, Beads, slop AI browsers, and AI-generated PRs flooding overwhelmed maintainers. I don't think we're ready for our new powers we're wielding. lucumr.pocoo.org/2026/1/18/ag...
ah, we traded places, I was on the roof of bracken house at the time! still gutted to have missed your debut at the expo!
I am astonished that this never came up! Looking forward to your next soundcloud drop.
sorry WHAT
set a bed-time alarm
Weeknotes for week 2, 2026: On climbing, drilling and interesting facts about drinking habits
alexwilson.tech/cont...
#weeknotes
the more interesting question is: where exactly has the money gone?
A table showing the performance of AI writing detectors (GPTZero, Originality, Pangram, and Roberta). The first 3 are commercial, the latter is open source. All of the commercial offerings achieve an accuracy of >95% on average. Roberta performs substantially worse with an accuracy of 10-20%. Raw table follows. Table 2: Detector Performance by Genre and Model: 4-Mean Model Genre GPTZero Originality Pangram RoBERTa GPT-4.1 amazon review 0.9675 0.9811 0.9642 0.1153 GPT-4.1 blog 0.9950 0.9722 0.9850 0.1090 GPT-4.1 news 0.9864 0.9982 0.9964 0.0185 GPT-4.1 novel 0.9960 0.7086 1.0000 0.0076 GPT-4.1 restaurant review 0.9700 0.9801 0.8859 0.0800 GPT-4.1 resume 1.0000 0.9994 1.0000 0.0191 Claude Opus 4 amazon review 0.9250 0.8984 0.9279 0.1990 Claude Opus 4 blog 0.9650 0.6869 0.9599 0.1078 Claude Opus 4 news 0.9733 0.9491 0.9872 0.0418 Claude Opus 4 novel 0.9630 0.4175 1.0000 0.0132 Claude Opus 4 restaurant review 0.9380 0.7750 0.8529 0.0923 Claude Opus 4 resume 1.0000 0.9759 1.0000 0.0193 Claude Sonnet 4 amazon review 0.9550 0.9744 0.9396 0.1936 Claude Sonnet 4 blog 0.9750 0.7509 0.9897 0.1244 Claude Sonnet 4 news 0.9900 0.9873 0.9933 0.0350 Claude Sonnet 4 novel 0.9900 0.7028 1.0000 0.0132 Claude Sonnet 4 restaurant review 0.9700 0.9048 0.8878 0.1105 Claude Sonnet 4 resume 1.0000 0.9984 1.0000 0.0193 Gemini 2.0 Flash amazon review 0.9250 0.9506 0.9128 0.1739 Gemini 2.0 Flash blog 0.9500 0.9023 0.9477 0.1025 Gemini 2.0 Flash news 0.9800 0.9933 0.9898 0.0071 Gemini 2.0 Flash novel 0.9950 0.9731 1.0000 0.0029 Gemini 2.0 Flash restaurant review 0.9200 0.8728 0.8049 0.0783 Gemini 2.0 Flash resume 1.0000 0.9993 1.0000 0.0140
AI writing detection is getting _really_ good these days. I hadn’t been paying attention to this part of the industry and I’m surprised (in a good way) to see how good it has gotten!
papers.ssrn.com/sol3/papers....