Owen Lacey

Owen Lacey

@owenlacey.dev

Developer / Maths nerd / Attention seeker ๐Ÿ“ Leeds, UK

378 followers236 following288 posts

Voyager update approx 1 week in: 1. Massive learning curve - I feel like I'm rewiring my brain 2. ZSA customer support is brilliant - I've been using typ.ing either side of abusing claude No idea what my wpm was before this but still feels slower, and I'm very tired at the end of the day.

It's a small thing, but feels like a milestone: I just closed my first issue raised against one of my open source projects. Feels good

Every single post of @matthen.com is a certified banger

Matt Henderson@matthen.com

Which Pokรฉmon does a compression algorithm think are most similar? bzip pairs Gengar with Haunter, Magikarp with Gyarados and Pikachu with Jolteon. 13 different Pokรฉmon get paired with Marowak- it's a very average sprite..

Two weeks later and I'm never going back

Owen Lacey@owenlacey.dev

TIL you can set a limit for YouTube shorts - it's always pissed me off how some apps become doomscrollified, and I'm extremely susceptible. I set it to 0 ๐Ÿ™

View full thread

I still don't understand why LLMs give time estimates for pieces of work. It always overshoots for me, and has no idea how autonomously it'll get delivered (nor do I at this stage). I'd much prefer it to give me it in terms of tokens.

Sanity check: which of these Rail Cube layouts bring you most joy? Are you able to articulate why? There is a correct answer

TIL you can set a limit for YouTube shorts - it's always pissed me off how some apps become doomscrollified, and I'm extremely susceptible. I set it to 0 ๐Ÿ™

I made something to generate blueprints for solves of my son's railcube, using every piece. How long until it finds a duplicate? (hint: it's at 2k an hasn't found one yet)

The estimate I'm happiest with is 28,339,256,149. How? Create a bunch of random tracks, and make note of the proportion of them that were valid. About 1 in 214,000,000 were valid, so multiplied by the entire problem set gives us this number. My solver is almost at 1M - I think I might cancel it ๐Ÿ˜…

I made something to generate blueprints for solves of my son's railcube, using every piece. How long until it finds a duplicate? (hint: it's at 2k an hasn't found one yet)

Found one - 10,000 solutions in. Solver found IOXSLIIOILLIOIISXOISRSSSSSRSSSSSRSRL, which is RSSSSSRSSSSSRSRLIOXSLIIOILLIOIISXOIS shifted 16 chars

I made something to generate blueprints for solves of my son's railcube, using every piece. How long until it finds a duplicate? (hint: it's at 2k an hasn't found one yet)

Best answer I can come up with is somewhere between 4000 (where the solver has got to) and 10333147966386144929 666651337523200000000 (35!) - help

I made something to generate blueprints for solves of my son's railcube, using every piece. How long until it finds a duplicate? (hint: it's at 2k an hasn't found one yet)

I took this out for a spin today and it was decent ๐Ÿ™Œ I did a before/after test on my current work project and it was able to get to the correct answer in a third of the time using the WebMCP tools I exposed

Owen Lacey reposted

To tell if a maze is solvable, just hang it by its corners If it tears into pieces, youโ€™ve found a solution

AI certs these days: "An SME is an AI skeptic. What's the most appropriate action to take?" 1. Call them a woke snowflake 2. Agree with them: AI is stupid 3. Do the middle finger wind up crank thing 4. Listen to them, show them quick-wins that will make their day-to-day easier