Sabitlenmiş Tweet
mj
5.5K posts
mj
@_mjmeyer
@redwoodjs core team • leveling up devs, indie hackers & my foundlings 👧👶👶 • aka the Dadalorian • digital gardening at https://t.co/NdvLQ2M1Or 🌱
frontend devs go here 👉 Katılım Ekim 2022
896 Takip Edilen1K Takipçiler
@chriszeuch i don’t think so.
there are many more people trying to build with ai not getting the results they were hoping for
unknown unknowns etc.
English
i'm curious if you've looked into observational memory? i came across it from the @mastra team, and have been experimenting with it on a side project. don't have hard stats, but i've been going on a session for what feels like 800K+ tokens (obviously compacting over time) and it hasn't felt like it's gotten dumber.
but not working on anything groundbreaking. so could be a false positive.
using this pi-extension: github.com/GitHubFoxy/pi-…
mastra.ai/blog/observati…
English
Adding opus 1m support meant we revised the context warning in humanlayer from “40%” to “100k tokens”
Matt Pocock@mattpocockuk
Doing some experiments today with Opus 4.6's 1M context window. Trying to push coding sessions deep into what I would consider the 'dumb zone' of SOTA models: >100K tokens. The drop-off in quality is really noticeable. Dumber decisions, worse code, worse instruction-following. Don't treat 1M context window any differently. It's still 100K of smart, and 900K of dumb.
English
evlog now has a @vite_js plugin
Add it to your vite.config and get:
• Auto-initialization (zero setup code)
• Debug stripping at build time
• Source location injection
• Auto-imports for evlog symbols
• Client logger support
Works with any Vite-based framework. SvelteKit, Hono, Express, you name it.
👉 dub.sh/better-log
English
this is great! 🙌
Hugo@hugorcd
evlog now has a @vite_js plugin Add it to your vite.config and get: • Auto-initialization (zero setup code) • Debug stripping at build time • Source location injection • Auto-imports for evlog symbols • Client logger support Works with any Vite-based framework. SvelteKit, Hono, Express, you name it. 👉 dub.sh/better-log
English
@wesbos @leerob @ephraimduncan been doing this in pi for the last month using this extension. should give you a good idea... github.com/sudosubin/pi-f…
English
You can now use Composer 2 in OpenCode with your Cursor Subscription
Cursor@cursor_ai
Composer 2 is now available in Cursor.
English
i will happily work for someone that thinks this way ngl
i seriously dislike fighting for more tokens
that or i need to make my own plan. i would prefer to not have to worry about finances on top of everything though
inb4 wouldn't we all. i know. i can dream can't i!?
TFTC@TFTC21
Jensen Huang: "If that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed. This is no different than a chip designer who says 'I'm just going to use paper and pencil. I don't think I'm going to need any CAD tools.'"
English
@GeoffreyHuntley oh man! for work or pleasure? would love to meet up for a coffee or something
English
@nrupatunga1987 @plannotator no. it executes a script and loads the last response from your session. once you reply, it injects the response as a prompt no different than if you had to manually respond
English
@plannotator @_mjmeyer But does that not grow the context unnecessarily?
English
mj retweetledi
Still a terminal user? Ever wanted to give detailed feedback on a long message? ... /plannotator-last
Works for all agents.
Thanks @_mjmeyer for the feature request!
English
vibe coding
just keep shipping and never make eye contact with the exposed secrets, broken auth, or confused customers
unicorn status awaits
James Lucas@JamesLucasIT
In 1984 a German reporter in Paris explains how to cross Place de la Concorde: "Walk steadily across and never look at the drivers. If you do, they'll think you've seen them and won't stop."
English
@kinglycrow love this. I've been toying around with a similar skills-based approach. Often ask: what do I need to know about the changes, and what code does each specific thing touch? (like a QA list)
x.com/plannotator/st…
plannotator@plannotator
Late night sneak peak - better QA review Have you ever asked an agent: "What all do i need to review and/or manually test"?
English
lately i've been pretty interested in tools that help me digest and understand large amounts of changes from agents so i built a little thing 🧵
English
@_mjmeyer @plannotator @badlogicgames Ive been using it all day. its really useful. great idea I will try to ship tonight, cleaning some things up & making avail to other harnesses like Pi
English
Underrated feature: annotate any file
@plannotator demoed in @badlogicgames pi. works in any harness as slash command or shell (!plannotator annotate file.md)
English
@backnotprop @plannotator @badlogicgames 🚀 ship it you legend! i can’t think of something better atm
English
English
@GeoffreyHuntley @dexhorthy good to know i’m not alone. i’ve tried pushing it but it just stops calling some tools eventually, and with long planning discussions it starts rambling like a mad man
English
just watched a video of @dexhorthy where he briefly mentioned @GeoffreyHuntley tells the agent to...
> execute the user's instructions then moo like a cow
and when it stops moo'ing like a cow, then you know the model is no longer paying attention.
i don't know if this is the smartest thing ever, but it kinda makes sense.
i've also been rofl'ing for the last 30min, so there's that. geoff bro, please tell us more! 🤣
English
@GeoffreyHuntley @dexhorthy what’s the vibe on the 1M context windows of opus / codex for you?
also, do you inject it as a hook or just randomly ask it to do it after it’s done working?
English
@_mjmeyer @dexhorthy it’s how i measure the dumb zone. been doing it since early sonnet days. always my litmus test when taming a new model.
English