@trq212 This problem isn’t that bad, but docs on telegram group setup would helpful (like group vs supergroup, finding group id…)
It’s great to plan a family trip, the Claude agent is running from our obsidian vault folder.
English
Rafael Zayas
63 posts
Rafael Zayas
@rafazaya
Tweeting to get better at tech, investments and product design. Portfolio Management and Trading @ Vident https://t.co/pckYE52j96
North NJ Katılım Mayıs 2023
443 Takip Edilen114 Takipçiler
setting this up is so fun, need to figure out a way to stream the dev process without getting instantly owned
Thariq@trq212
English
English
Gonna code for like an hour before bed. What are the biggest issues in T3 Code I should fix right now?
English
@AccountMus629 @theo Hmm I’m also on niri/cachyos and haven’t had this problem. My guess is you can set up an xdg var so it knows which file dialog to use.
My different problem is that links don’t click through for me, and I’m trying to sort that out, but I think it’s the same class of problem.
English
@mattpocockuk Which is kind of interesting right? How we mentally embed details of implementation in our “gray box” understanding.
I suspect the next generation will bear less cognitive load per project, and us implementors of old may be outgunned.
English
@rafazaya Yeah I'm wondering if it's even 3x
It's unbelievably draining trying to keep all this in your head
That's what a bottleneck feels like
English
A wonderful description of Cognitive Debt:
"Even if agents produce code that could be easy to understand, the humans involved may have simply lost the plot and may not understand what the program is supposed to do"
I have thoughts on how to fix this.
margaretstorey.com/blog/2026/02/0…
English
@mattpocockuk An interesting question then is what is the AI leverage one can achieve, i.e. how many more projects can you oversee as “gray box” vs deeper understanding. Is it 10x leverage? I don’t think it’s 1000x or anything like that based on how mentally tired I am each day.
English
However, you need to be disciplined in keeping a shared understanding of these across your team.
This is an open question. But some ideas:
- Pair program
- Get AI to do weekly reports on the main interfaces in the program
- Do RFC's for interface changes
English
The scorched earth AI take that “your org rarely has good ideas. ideas being expensive to implement was actually helping” is so hot, and so obvious, that I’m seething with jealousy I didn’t think of it first.
dax@thdxr
everyone's talking about their teams like they were at the peak of efficiency and bottlenecked by ability to produce code here's what things actually look like - your org rarely has good ideas. ideas being expensive to implement was actually helping - majority of workers have no reason to be super motivated, they want to do their 9-5 and get back to their life - they're not using AI to be 10x more effective they're using it to churn out their tasks with less energy spend - the 2 people on your team that actually tried are now flattened by the slop code everyone is producing, they will quit soon - even when you produce work faster you're still bottlenecked by bureaucracy and the dozen other realities of shipping something real - your CFO is like what do you mean each engineer now costs $2000 extra per month in LLM bills
English
soap is crazy because it’s like claude code for your hands
English
@Glas109 @theo @YaLTeRz I was going to suggest dms after starting to read your tweet… CachyOS ships Noctalia shell by default now, which a lot of people like. Also, most AI is good at setting up niri configs, especially if you link the wiki, so if you can say what you want you can probably make it.
English
If you can afford a $200/month Claude Code subscription, you can afford $100/month to someone who made something you love.
Thank you @YaLTeRz 🫡
Theo - t3.gg@theo
The Niri window manager feels like exactly what I was looking for. Holy shit this is good.
English
English
English
@compuficial @theo I have the gen 12 of this, and it’s really good, but I think the trackpad would be a downgrade from the MacBook. But maybe that’s just a Linux driver thing. I’m watching this thread for other refs though.
English
The Niri window manager feels like exactly what I was looking for. Holy shit this is good.
Theo - t3.gg@theo
English
Also spoiler, niri has one-shot me.
Saw this reply, "Vertical" and "horizontal", looked into niri and...yeah fuck me I'm currently tweeting this from Linux.
x.com/rafazaya/statu…
Rafael Zayas@rafazaya
@DanePoyzer @theo I started using niri just before I got into claude code, but that way of scrolling horizontally within a project, and vertically between projects fits my mental model quite well. The annoying thing was to create browser profiles to spawn new browser windows in each workspace.
English
@DanePoyzer @theo I started using niri just before I got into claude code, but that way of scrolling horizontally within a project, and vertically between projects fits my mental model quite well. The annoying thing was to create browser profiles to spawn new browser windows in each workspace.
English
@theo I use a virtual desktop for each
I use tmux, with a window for each project split into panes across server/shell/agent. I name the panes so that when I get a ping it’s clear which
I match local db/server ports eg proj 1 is 8081/3001 and so on
Terminal/browser/ide per desktop
English
@karpathy I’ll never forgive Google for shutting down Google Reader, it was nearly perfect, and the signal to noise of being online was SO much higher back then then.
English
Finding myself going back to RSS/Atom feeds a lot more recently. There's a lot more higher quality longform and a lot less slop intended to provoke. Any product that happens to look a bit different today but that has fundamentally the same incentive structures will eventually converge to the same black hole at the center of gravity well.
We should bring back RSS - it's open, pervasive, hackable.
Download a client, e.g. NetNewsWire (or vibe code one)
Cold start: example of getting off the ground, here is a list of 92 RSS feeds of blogs that were most popular on HN in 2025:
gist.github.com/emschwartz/e6d…
Works great and you will lose a lot fewer brain cells.
I don't know, something has to change.
English
Rafael Zayas retweetledi
Who's exactly serving whom in this scenario?
NetworkChuck@NetworkChuck
My server has a phone number now. I can call it from ANYWHERE (even a payphone in the middle of nowhere with zero internet) and I can talk to Claude Code. But that's not the crazy part. My server can call ME. When something breaks, it picks up the phone and tells me about it. Check it out: youtu.be/cT22fTzotYc @3CX
English
@thdxr Is there any info on how the usage is licensed? I ask because the copilot cli tool is under the pre-release software license, which has much less data protection than the regular enterprise copilot usage. @thdxr @opencode @jaredpalmer
English
the pool of people who have access to copilot is huge
every enterprise microsoft customer has already bought it for their employees
it's pretty amazing they're supporting OpenCode when other company is aggressively doing the opposite
OpenCode@opencode
OpenCode can now officially be used with your Github Copilot subscription with the $39 pro+ subscription you get access to the best coding models wonderful to see them support open source and user choice of tooling in this way
English
@opencode Is there any info on how the usage is licensed? I ask because the copilot cli tool is under the pre-release software license, which has much less data protection than the regular enterprise copilot usage.
English
@jaredpalmer @knotbin @thdxr I'd love to be able to use opencode connected to github copilot! Would the data policies be the same as accessing via VS Code?
English
This is awesome, I apologize to the team in advance for annoying them about considering this approach for all new tests.
Charlie Marsh@charliermarsh
The ty test suite is "written" in Markdown. Every code block here gets evaluated, with the comments representing expectations. We have almost 300 of these files that effectively read as detailed documentation for how ty behaves and how the Python typing spec and runtime work.
English
@theo Wait, does search not work? Is its knowledge recent and off the rip?
English
@theo Im only a few minutes into using it, but using in t3 chat with search grounding is very very good
English
What are the coolest things you've seen people do with the Horizon models?
English