Sabitlenmiş Tweet
Palash Bansal
613 posts

Palash Bansal
@repalash
building stuff https://t.co/jYKSNAYwG7 https://t.co/uFsnfqIINx https://t.co/CAtPuwbXAH
San Francisco, CA Katılım Eylül 2012
513 Takip Edilen3.2K Takipçiler

It was made for humans, handles things like scrolling, terminal size etc which are not required for ai tool use. In most cases because of these differences it just doesnt work, then you need more special flags to enable/disable this.
Also, there is no proper standard or structure to the output. People use hacks like add support for --json --yes, but its all custom. even claude cli doesnt provide a proper way for everything you can do interactively, like hooks and all.
session management in cli processes also has to be added separately (mcp should have something for this, not sure if it does though).
I agree mcp was a mistake, but cli is not the solution for that
English

@repalash CLI is a standard since 90 at least, --help is the context then the LLM should just use the right commands
English

Every MCP could be a CLI or a cURL call
Pietro Schirano@skirano
MCP was a mistake. Long live CLIs.
English

Not all cli can be interacted with in the same way and they dont follow the same protocol. It was made to make it consistent, once protocol is made, it can be followed by http servers, or other non-bash-like shell without any change on the other side. This seems better to finetune on than cli tool calls.
English

@minjunesh Can it publish my macos app(3dviewer.xyz) also to the appstore?
and port the 3d quicklook to ios while its at it
English
Palash Bansal retweetledi

Holy shit Claude passed App Store review submission in one shot. I had it make the demo app "ParkSaver" using blitz.dev and submit it in one session to record promo videos for Blitz.


English

350-400 hours into OpenClaw over the last 33 days non-stop, no days off...I'm ready to quit.
My openclaw is fucking lost in the weeds every day today and it's driving me nuts.
Basic shit. I asked it to use GitHub. it has a GitHub skill. We have a GitHub SOP.
I can see it's thinking process about using skills, then narrating how the skill doesn't exist, then going and inventing ways to retrieve the capability to use GitHub from the internet.
I tell it to look in the openclaw docs for the proper skill path, it says "oops my bad, yeah it was there after all."
This is ChatGPT 5.4 with extra high thinking turned on.
I ask it to diagnose the problem only, so it goes and sees the system prompt is telling it to look at the wrong place, and it goes to GitHub and opens a GitHub issue about this 'bug' without even asking me.
What the actual fuck. 3 hours on a Sunday of trying to rewire the brain of my openclaw to do default-behaviour.
This thing such a productivity suck & mental poison.
I can't do anything useful or positive with OpenClaw because I'm nonstop fighting fires in the engine room.
I'm thinking about giving up.
English

@cmuratori @bulletproof_ian @hardwarecanucks I have experienced quite a few crashes with windows 11 recently, and I didn't see any screen blue or black, I think they are just removing it makes them look bad.
English

@bulletproof_ian @hardwarecanucks Get with the times, boomer! It's not blue anymore. It's black: theverge.com/news/692648/mi…
English
Palash Bansal retweetledi

Coding iOS apps is easy now, but submitting to the App Store is still a huge pain.
Blitz solves this by giving Claude Code tools to build, test, and submit your app all from one place.
Just ask Claude to "submit my app to the app store"
👉blitz.dev


English

@pierskicks @threejs What's WebGPU got to do with it?
Only people who have a discovery layer to sell say that it's the missing piece (there are like 50 more pieces).
It's not even one of the biggest issues, io games etc never needed it.
English

@martin_casado No bro, you just have to prompt harder and connect the right MCPs
English

@youyuxi @fernandorojo Tell that to Claude when it's trying to sed random bytes from a 5mb built library with no new lines.
It hallucinates/makes up more stuff when reading from bundled code than a codebase with proper structure and file names that it can navigate through
English

@fernandorojo Agents can read compiled / bundled code just as well. They don’t really care. In fact it’s better to be reading code that is actually being executed than source code (which might go through arbitrary transforms / indirections before becoming the built artifact)
English

@samhogan I can't tell if this is a joke or serious. You know postgres also writes data in files right, this is just adding extra steps
English

What if a codebase was actually stored in Postgres and agents directly modified files by reading/writing to the DB?
Code velocity has increased 3-5x. This will undoubtedly continue. PR review has already become a bottleneck for high output teams.
Codebase checked-out on filesystem seems like a terrible primitive when you have 10-100-1000 agents writing code.
Code is now high velocity data and should be modeled at such. Bare minimum, we need write-level atomicity and better coordination across agents, better synchronization primitives for subscribing to codebase state changes and real-time time file-level code lint/fmt/review.
The current ~20 year old paradigm of git checkout/branch/push/pr/review/rebase ended Jan 2026. We need an entirely new foundational system for writing code if we’re really going to keep pace with scale laws.
English








