Magnus Strømseng

@shivamhwp Bro, the point is not having to use their UI, if you are required to have Outlook by work/school.

You think i would've been on Cron, sorry Notion Calendar if i wanted to use outlook ???

@kitlangton @thdxr @opencode Oooh been waiting on this for some time!

taking advantage of the new inline 𝚘𝚙𝚎𝚗𝚝𝚞𝚒 rendering system for some @opencode micro-animations ⭐

@devinjameson Wow cool! Didn’t know you had implemented this haha

@stroemseng Totally agree. foldkit.dev/ai/mcp

People are talking about how agents should be able to access everything. We should automatically create agent APIs from pages that use state machines, and therefore avoid agent UI interaction.

@RhysSullivan @zeeg Yeah, opencode repo has been a big inspo for me

@zeeg now that the LLMs are writing my code i like effect a lot more, just give it some good references like the opencode repo, executor repo, and effect-smol source and you should get good outputs without too much pain

should I rewrite my project in Effect or will i just get annoyed

@kennethnym @geerlingguy has a great video comparing many different ones!

speaking of remote kvm (comet pro is great): pikvm is an open source ipkvm stack that lets you turn a raspberry pi into a remote kvm! there are also alternatives like tinypilot which offers the same thing. its a great weekend project! tempted to build one myself..

@alan__rice Fold proteins yeah, fold clothing no.

I AM HOPEFUL BUT SKEPTICAL CAN IT REALLY FOLD???

@f4micom These were all her?! Wow yeah

between chatcontrol, age verification bullshit and now the tariffs i genuinely have lost all respect for ursula ngl

@zeeg Yes please!

Should I write up a blog post about Junior or are super-capable multiplayer agents played out at this point? Pretty sure our harness is a lot more advanced than most out there and has some interesting technical decisions. junior.sentry.dev for context

@zebassembly React scan is really good. The fact that it's kinda needed is really bad. There are more guns than feet at this point.

@omarsar0 Huh, interesting approach. Gotta try this out

If you use LLM-as-judge, this one is worth reading. (bookmark it) It's actually one of the most effective ways to use LLM-as-a-Judge for evals. Holistic judge scores hide both their reasoning and their ceiling effects. BINEVAL decomposes each evaluation criterion into atomic yes-or-no questions, answers each independently per output, then aggregates the verdicts into calibrated multi-dimensional scores. Every question-level verdict is inspectable, so you can diagnose exactly why an output scored low, and the same verdicts feed straight back as targeted prompt-improvement signal. Across SummEval, Topical-Chat, and QAGS, it matches or beats UniEval and G-Eval, training-free, with especially strong results on factual consistency. Paper: arxiv.org/abs/2606.27226 Learn to build effective AI agents in our academy: academy.dair.ai

@mattzcarey @RhysSullivan Isn’t this just a REPL then? Like the RLM paper? Or am I misunderstanding?

@RhysSullivan Stateful codemode

tbh i wonder if we're overindexed on codemode the way we were with standard tool calling & then bash after that code mode is the best approach so far, i think you can call anything from it and you get return types, chaining, filtering, but wonder if there's something more

Universities should teach more software architecture, maintainability, and system design. With LLMs generating most code, the real challenge is building systems that stay clear, scalable, and maintainable. My university barely covered this, I had to learn it on my own.

@pnpmjs 👀

Very early sneak peek to pnpr - the pnpm registry: pnpm.io/pnpr/

@leo_trapani Hey, I’m working on this combo as well!

The more I dig into flue's code, the more I see how good of a job was done I've been trying to build something like flue's model on cloudflare (Durable Object, durable streams protocol) but Effect and Alchemy native, but I think I'm not going to implement all the recovery stuff (is the worker crushes, we can restart MID run) That's just beauty @FredKSchott

@hagaetc Yeah, Handelsbanken ain’t exactly much better, but hey I get the best interest with them through Tekna 🤷‍♂️

It's 2026 and Norway's biggest bank DNB will not show you your IBAN natively with your account, have a separate annoying app to generate your IBAN, will not let you copy your account number from the UI *AND* will not let you paste an account number into the the damn generator🫠

Can’t believe they made @bryan_johnson the villain in Vox Machina season 4 (He’s trying to cure death)

@RhysSullivan Shouldn’t you leave some alphabet cookies instead, so it can consume some tokens?

make sure to leave weights and matricies out for anthropic tonight incase fable comes

Valve - “starting with the SteamOS 3.8 release, you can put together your own Steam Machine using whatever PC parts you want.” theverge.com/games/953411/v…

@theo Wow, cool product. They open sourced the software, and they have guides to build your own hardware!

Set up my new trmnl to show my monthly token usage. Not sure if this is good for me 💀