on the topic of using Claude Desktop for personal projects, happy to report that Claude Design now actually is quite good indeed - last time I tested it it was when it was first rolled out in beta and it's night and then since then

i use LLMs via CLI at work (pi for the win) so despite being aware of such a thing, i was pleasantly surprised when i told Opus 5 in the claude dsk app about a mobile specific bug for the website i cooked a while ago it just went "well i have the iOS sim, might as well" 👏

interesting - first time I'm seeing an AI being "self-aware" ahead of time about its token/usage quota (opus 5, mid) 👀

quick update from two "5hrs limits" later: Fable seems to have the exact same issue as every other model - once it funnel itself into an idea, it doesn't get out of it no matter what I said. I literally told it to throw away all the code and restart again, and yet he stuck to it

im gonna be far from my personal laptop for a few days so decided to crank things up to max and just give it a big chonky task to see what it'd end up with 🍿 also just realised that it's the first time i'm pushing ANY model above xthink (always got "scared" by cost/consumption)

and to be fair, also last week Jetbrains announced the same for Kotlin: blog.jetbrains.com/kotlin/2026... ...but the models chart is so outdated 🫠 🔗 kotlinlang.org/benchmark/

ICYMI last week the Android team revamped their eval board for Android and ngl I'm somewhat surprised how much better than opus 4.8 Sonnet 5 is. (it's still missing the latest 5.6 gpt rounds tho) 🔗 developer.android.com/bench && announcement: android-developers.googleblog.com/2026/07/and...

a bit late on this one but i immediately installed on my personal machine and will keep it there forever: MDN is great and i'm sure having the agents able to leverage it better/more directly is gonna make website building better overall 🔗 developer.mozilla.org/en-US/blog/...

been hearing about headroom lately - have you used it? thoughts? 🍿 🔗 github.com/chopratejas...

someone's feeling scottish today 🤖

from what i've read, tbh, I think that we'll see mythos & fable back up sooner than we think. BUT while we wait I want to put a spotlight on one of my fav benchmarks out there: minebench. The author always does great breakdowns, like this one: www.reddit.com/r/singulari...

yes yes I know now everyone will just talk about Mythos and Fable the whole next 24hrs but did anyone poke at the Xcode 27 MCP referenced here? couldn't find dedicated docs 😭 I wanna know if it's decent! The Android CLI is so good I want Apple to build the same thing for iOS.

finally took odysseus for a spin this morning 👀 with Gemma 4 it's mindblowing that this is all running locally on a relatively small ram laptop (24)* as an #oss project it's still rough around the edges but has a ton of potential - amazing amazing stuff

small 💎 to clean up your feed going into the weekend 🔗 miniscript.org/MiniMicro/i...

PewDiePie's journey into tech and #opensource has reached its peak and I'm all here for it and the fact that both him and Antirez have been heavily using Deepseek v4 really talks volumes about the impact of that project cant wait to see how it will evolve

it's the end of the month so you know what that means... i have 36hrs to burn through that 50% 😂 🔥🧑‍💻🔥🧑‍💻🔥🧑‍💻🔥🧑‍💻🔥🧑‍💻🔥🧑‍💻🔥🧑‍💻🔥

gave a first quick run to the new opus 4.8 (xthink) on my personal machine via the Claude app - it ran for circa 3hrs and overall... it feels allright? my initial impression is that this is the "jump" we should have felt with 4.7. not sure if I would use it instead of gpt5.5 tho

🤦‍♂️

let's see how it is 🍿

lowkey the "✨Ask" feature on YT is one of the best AI integrations anywhere. probably uses the same brain as NotebookLM under the hood, great timestamps and breakdown. (I don't use it for vids like this one, i just needed an example to take 📸) (but this channel is peak)

bank holiday maintenance sweep on the good ol’ upgrade-helper repo. first real #oss upkeep in ages, and my first time using agents for it. they made the grindy bits way faster - all issues and all PRs are commented to/acted on. really, ai can be a solid tool for oss.

Very interesting idea to put directly in the copilot CLI - I do something similar myself when I need it already, but I can see the value in using it to leverage a smaller model more efficiently. I might try it out to see how it goes 🔗 github.blog/ai-and-ml/g...

there's so much to unpack in this announcement but i want to put a small spotlight on one of the things that people will care less for: 🔗 www.anthropic.com/glasswing

im really trying to play around with all the main agents to keep myself flexible but gemini you really need to step it up, can't have this level of stalling at random times :eyeroll:

i know i've already mentioned but for websites the chrome devtools mcp is so good

very useful insight from The Pragmatic Engineer newsletter about what happened with Blocks 👀 🔗 newsletter.pragmaticengineer.com/p/the-pulse...

been catching up with the AI world and was reading Simon Willison’s Weblog - in particular his Agentic Engineering Patterns section: simonwillison.net/guides/agen... decided to read through it and extract some improvements for to agentsparty.dev ... quite a beefy addition 🍿

recently I tried multi-agent stuff hands-on and realized it’s basically d&d: you've got a game master (main agent) directing a party of sub-agents, each with their own special abilities to tackle tasks naturally, I leaned into the theme hard and built the whole thing like so 🧙‍♂️

soooOOOO I've been nerding out on multi-agent orchestration lately and ended up building something silly that I kinda want to share 🎲🤓 (beware, wall of text incoming 🧵) agentsparty.dev 👀

might be 🧑‍🍳 here 👀👀👀