Aidan Harding @aidanharding.eurosky.social · 15hExmoor is at its best for #trailrunning when the skies are big and the sun is low for autumn 070 7
Reposted by Aidan HardingRick Burin @rickburin.bsky.social · 16/10/2026Reminded today of this all-timer from the early days of The Internet 13549129 549
Aidan Harding @aidanharding.eurosky.social · 15/10/2026www.smbc-comics.com/comic/2013-0...smbc-comics.comSaturday Morning Breakfast Cereal - 2013-02-01Saturday Morning Breakfast Cereal - 2013-02-01 030 3
Reposted by Aidan Hardingp(Dulany) @dulanyw.bsky.social · 13/10/2026Interviewer: can you explain this gap in your CV? Person who wrote the layout engine for Microsoft Word: absolutely not 357293 572
Reposted by Aidan Hardingamos @fasterthanli.me · 12/10/2026- Can you pass the salt? - Passing the salt is real work. Probably about a day's worth. Given we're on track to finish lunch today, I probably would budget that for next quarter. - Just pass the damn salt *passes salt* - Okay, you win this one. It actually took three seconds. Listening and learning. 415110 151
Aidan Harding @aidanharding.eurosky.social · 12/10/2026Ironic how the running is hardest to fit in when you need it most. Good job getting out there! 110 1
Aidan Harding @aidanharding.eurosky.social · 11/10/2026Getting back on the #trailrunning after too much work and too much illness recently. Luckily for Aurora and me, autumn is still showing off, even if we did get wet paws #northdevon 1140 14
Reposted by Aidan HardingThorne 🌸 - Play Cozysters! 🎢 @ens0.me · 08/10/2026Software engineering is now about being able to make quality decisions around architecture, design, and technology. The actual engineering all along, the code just was in the way. 41067 106
Aidan Harding @aidanharding.eurosky.social · 06/10/2026Them: AI stops you learning anything. Me: Needing to remind myself of how a specific data system works, gets Claude to generate an interactive quiz/nature-guide with every fact cited from the latest docs and even local LLM free-text questions so I can use it on a train with spotty data connection 151 5
Reposted by Aidan Hardingsocketwench @deninet.com · 06/10/2026Still thinking about a comment a read in response to "fast, cheap, or good, pick two" that responded with, "I never expected companies to pick fast twice." 012431 124
Aidan Harding @aidanharding.eurosky.social · 04/10/2026Beautiful autumn run from home this morning with low sun and chestnuts. It was beautiful to see the steam coming from Aurora's paws as she ran up the hill into the sun #trailrunning #northdevon 050 5
Reposted by Aidan Hardingren faire @lathrys.at · 03/10/2026the problem with ai for the kind of person with my kind of brain is not that it offloads hard cognitive work, per se, but rather it enables a certain disposition to run 24/7 best summarized as: 330339 303
Aidan Harding @aidanharding.eurosky.social · 30/09/2026Disagree. PhD opining outside their area of expertise is the warning sign. Displaying your credentials is fine. Some people get pre-judged as 'non-technical' or 'non-expert' without such displays. 000 0
Reposted by Aidan HardingKiran @kirancodes.me · 24/09/2026"file"-system? buddy. I gotta claudesystem 91485 148
Aidan Harding @aidanharding.eurosky.social · 23/09/2026Although CS inclined would think that MinMaxing has the standard problem of optimising a single dimension at the expense of others 010 1
Aidan Harding @aidanharding.eurosky.social · 23/09/2026What 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 110 1
Aidan Harding @aidanharding.eurosky.social · 21/09/2026Fair point... 10x cheaper for the flash model, so then our evalutaion becomes a question of the quality we get on those models. I'll give them a try. I get that API prices are going to be more expensive. It just seems that the result is being locked in to Athropic and OpenAI by $$s. 000 0
Aidan Harding @aidanharding.eurosky.social · 21/09/2026This was via fireworks.ai on a serverless package. Are the frontier labs simply an order of magnitude cheaper?fireworks.aiOwn Your Specialized Intelligence | FireworksFireworks’ state of the art training and inference platform take you beyond the frontier, transforming open models into your specialized intelligence. 110 1
Aidan Harding @aidanharding.eurosky.social · 21/09/2026Last 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! 221 2
Aidan Harding @aidanharding.eurosky.social · 19/09/2026Snedding some trees and I thought the dog would like to come. The inevitable happened 030 3
Aidan Harding @aidanharding.eurosky.social · 18/09/2026Just two bros who like sitting on each other for cuddles and grooming #cats 0100 10
Reposted by Aidan HardingThe Unquiet Ghost of Andrew S. @shoutingboy.bsky.social · 18/09/2026important update 2254 25
Aidan Harding @aidanharding.eurosky.social · 18/09/2026I've thought that, but are we underestimating AI's chance of swallowing architecture too on that timeline? Today, architecture is reserved for humans only because the AIs are bad at it. But that's just scale/capability. Product Owner seems like a more durable place for the border to land 030 3
Aidan Harding @aidanharding.eurosky.social · 17/09/2026I 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 010 1
Aidan Harding @aidanharding.eurosky.social · 17/09/2026I ended up making my own setup for working like this, an open source claude skill using ideas from @dollspace.gay to do adversarial critique and refinement of the spec. github.com/aquivalabs/s... It's worked pretty well with our team. Not perfect, but what is?github.com 130 3
Reposted by Aidan HardingP(aul) Frazee @pfrazee.com · 17/09/2026Software dev is all the 2nd 80% now 1624932 249
Aidan Harding @aidanharding.eurosky.social · 16/09/2026We've ended up reviewing specs and docs as humans. Code largely (but not exclusively) via AI. It's imperfect, but pretty good. At least over the duration of a few months. It's close to what you're getting at. Specs are elaborate prompts and we aim to make them hold the review-worthy stuff 130 3
Aidan Harding @aidanharding.eurosky.social · 16/09/2026In a team! Absolutely. It doesn't seem to be discussed anywhere near as much as individual productivity. Also over real duration and software lifecycle. I've been doing this and plan to blog in our methods soon. If you have read good stuff, please share 140 4
Reposted by Aidan HardingLorin Hochstein @norootcause.surfingcomplexity.com · 15/09/2026Reminder that technologists have no special ability to predict the impact of the technologies they build. 1152 15
Reposted by Aidan HardingClaire Cullingworth @ccullingworth.bsky.social · 14/09/2026This made me smile, attached to an ordinary cast iron drainpipe on a house in Durham. #TinyJoys 231971413 2K
Reposted by Aidan HardingJess Miers 🦝 @jmiers230.bsky.social · 13/09/2026AI 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. 466290 662
Aidan Harding @aidanharding.eurosky.social · 13/09/2026@originstorypodcast.bsky.social's most recent episode is great in how this madness came to be 020 2
Aidan Harding @aidanharding.eurosky.social · 09/09/2026Advances in AI can only be understood backwards, but they must be lived forwards - not-quite Kierkegaard 021 2
Reposted by Aidan HardingSingularity's Bounty e/CC @catblanketflower.yuwakisa.com · 08/09/2026Chinese 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 2396 39
Reposted by Aidan HardingBrian Bilston @brianbilston.bsky.social · 07/09/2026Today’s poem is called ‘The Grammar Police’. 1131378 313
Reposted by Aidan HardingDorian Lynskey @dorianlynskey.bsky.social · 07/09/2026Street fascism doesn't start with beatings. It starts with an intimidating show of force. To call that "peaceful" is absurd. 1031565 315
Aidan Harding @aidanharding.eurosky.social · 06/09/2026But reading code is increasingly becoming a narrow way of deciding if it's good enough. I care about architecure, shape, and correctness. None of those require reading all the code. AI gives us more test options, more architectural documentation (especially if you build from specs), etc. 1110 11
Aidan Harding @aidanharding.eurosky.social · 06/09/2026😀 If I'm ever 50% about going for a run, I know someone a dog who is going to be 100% up for it! 110 1
Aidan Harding @aidanharding.eurosky.social · 06/09/2026I think AI leaves only the hardest and most interesting parts to the engineer. Which is invigorating and exposing all at once. If you were a bad engineer(ing organisation) before, you're going to be turbo-bad with AI. If were good, you're going to have the chance to soar. 3474 47
Aidan Harding @aidanharding.eurosky.social · 06/09/2026Running round the woods with Aurora. The rain has come back and made Devon lush and green again! #trailrunning #dogs 1141 14
Reposted by Aidan HardingBryan Cantrill @bcantrill.bsky.social · 06/12/2025Your intellectual fly is open bcantrill.dtrace.org/2025/12/05/y...bcantrill.dtrace.orgYour intellectual fly is open | The Observation Deck 815834 158
Reposted by Aidan HardingP(aul) Frazee @pfrazee.com · 06/09/2026We should stop thinking about software as anything other than getting the computer to do what we want it to do for us 2536743 367
Aidan Harding @aidanharding.eurosky.social · 05/09/2026This 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...stratechery.comWho’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. 130 3
Reposted by Aidan Hardingsrrrse @without.boats · 04/09/2026Every 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. 18010 80
Reposted by Aidan HardingWill Quinn @willquinnart.bsky.social · 04/09/2026Daily bunny no.3431 finds a friend (or foe?) 202518490 2.5K
Aidan Harding @aidanharding.eurosky.social · 01/09/2026Part 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 000 0