Aidan Harding
@aidanharding.eurosky.social
Architect at Aquiva Labs on Salesforce and cloudy things. Also trail running and country living in N Devon
What an amazing @originstorypodcast.bsky.social on John Rawls. "What if you had skin in the game but didn't know what colour your skin was going to be?" Also the questioning of meritocracy which seems so naturally true, but the isn't. Great stuff
Last week, I tried open weights models for coding compared to my existing method of a Claude Team Premium account. And I'm missing something, or the cost per token difference makes open weights unusable for me. Reviewing one PR on GLM-5.3 cost more than a day of tokens with Claude!
Snedding some trees and I thought the dog would like to come. The inevitable happened
I just read "sprint" in a document and my inner voice read it as "s-print" like sprintf in C, which is language I've barely touched in 20 years. Maybe I need a /clear on my own context today
Reminder that technologists have no special ability to predict the impact of the technologies they build.
AI can now detect, automate, and chain together exploits faster than any human hacker can. That is significant. Which means that the system vulnerabilities once considered not so low hanging fruit are actually very low hanging.
Advances in AI can only be understood backwards, but they must be lived forwards - not-quite Kierkegaard
Chinese Room only demonstrates that the operator doesn't need to understand. It doesn't say the operator *cannot* understand. The operator may still understand. If the operator decides to take a night course in Chinese to make his job easier, the room still works exactly the same
Street fascism doesn't start with beatings. It starts with an intimidating show of force. To call that "peaceful" is absurd.
Running round the woods with Aurora. The rain has come back and made Devon lush and green again! #trailrunning #dogs
We should stop thinking about software as anything other than getting the computer to do what we want it to do for us
This is really good perspective on open weights and the difference in economic model between conventional software (0 cost of goods) vs AI (costs on every inference). Although it does skip over cloud and SaaS being very much the latter camp. stratechery.com/2026/whos-af...
Who’s Afraid of Chinese Models?
Everyone is worried about Chinese models, but the frontier labs will be fine; we need to enable open U.S. alternatives.
stratechery.com
Every programmer knows Dijkstra’s “Testing shows the presence, not the absence of bugs” and yet every programmer goes about their day pretending it isn’t true.
srrrse@without.boats · last mo.A lot of professional programmers don’t even seem to know what a proof is or why they’d want one
Part 1 of realising we'd got too attached to Anthropic was checking with my immediate team on Opus 5, then wider team, then cross company. Everyone finds it slow, expensive and obtuse. Simply going back to 4.8 has helped loads. Long term, keeping options open with other models and harnesses
23 YEARS AGO TODAY, @mrweebl.bsky.social posted BADGERS to @Newgrounds.com
Nature fact: If you encounter a bear, make yourself look big From the Private Eye archive. Issue 1621, April 2024.
Quintessential #exmoor #trailrunning. The heather was blooming purple The clouds low, dumping both rain and sunshine at times. Seas of grass with hints of tracks. Great stuff
OMG, a while back, I added a section to Claude.md to look for some of the standard cognitive biases and it actually works. Sunk cost fallacy check in action: "We chose Rooms-in-scope on my incorrect claim that the data was already there. Worth asking whether we'd choose it again now."
When I used to write for The Guardian, every time I wrote the word “meow” the editors would change it to “miaow”. It used to really get on my wick. Don’t try to make my cats sound like your upper middle class London cats, you fuckers. My cats did not go to your posh cat schools. They say “meow”.