I made a contexto.me solver.
I was able to zero in on it using the GloVe 300d dataset.
Uses lemmas under the hood (runs, run, runing -> run), and then "just" find the word(s) where the guess is X distance away.
I think my best score before this was like 50.
This is exciting. The majority of my projects involve using Googles CP-SAT solver in some way. Today, I got a PoC js port working
Reinventing the wheel is underrated
me prompting claude "yeah if they've not set up rate limiting that's not our fault"
Current status:
There's also just six-degrees.app for you film buffs
This is one of those things you make that barely anyone notices, but gives me endless joy. Site is football.six-degrees.app/daily
Can confirm @samwho.dev is equally sound IRL
One thing they don't tell you about parenting - getting up at stupid-AM for a flight is an absolute pisstake compared to before
100 jumps game, where I got to 99 jumps and died
Pain.
I spent 2 months learning about quantization and am extremely proud of the post I've written about it. I think these are some of the nicest visuals I've ever made, and I love how this compression technique invented in 1898 is being used on the bleeding edge in 2026.
ngrok.com/blog/quantiz...
Cryptle medium 6/6
🟥⬜⬜⬛
🟥⬜⬛⬛
⬜⬜⬛⬛
🟥🟥⬜⬛
🟥⬜⬛⬛
🟥🟥🟥🟥
Took me longer than I care to admit
I'm at the "I'm a genius" stage of a side project. It can only get worse from here. In 24 hours I'll likely be in the fetal position.
I wonder how long it'll be before companies just put AI features in their usual branding. No more purple / rainbow gradients, we'll just expect them to exist as users.
Yes, or I would even come back to it at a later time (after it's usually caused me a few headaches)
I have a decent amount of success asking it "how does feature X work?". The process of me understanding it usually means I can make improvements here or there
Screenshot of WhatsApp's delete message modal. The 'delete for me' is circled
I bet you can't find a button that's pressed less than this
I need my own CLAUDE.md file to stop me from continually doing the same stupid shit
This is so cool!
What do LLMs see?
I wrote a lil' tool that extracts the attention matrices out of open models and creates this typing visual, with each token's opacity changing according to its average attention score as the prompt progresses. Dimmer words are considered less important to the model.
Despite my best efforts to avoid learning python, I'm a sucker for interactive stuff like this 😍
0byte.io/articles/pyt...
Having a little fun with some fragment shaders. Progressively quantising the word "quantisation" because I have no sense of subtlety.
This is gorgeous 😍
In graph theory, there are algorithms that find the shortest path between two nodes. I made one with pure CSS (including the graph drawing).
Drag the nodes, and the shortest path will update in real-time!
css-tip.com/graph-theory/
A fun demo powered by all the modern & cool CSS features 🤩
That's valid
If you're happy to use a fake one, ofcom maintains a list of thousands for use in film/TV
www.ofcom.org.uk/phones-and-b...
Also good for test data FWIW
Dear internet, stop making everything into a lasagna. Sincerely, everyone.
That's mad. I genuinely have a signed autograph from him. I cried on my first day of school when people called said "oh you're like Michael Owen" and I thought they were teasing me
Sorry, what?
I vibe coded a poc for something like it recently. It's not as swish, but I'll share the code with you!
