Prash Krishna
209 posts
Prash Krishna
@KrishnaPra6837
AI builder and PM. “You are absolutely right!” My views.
San Francisco, CA Katılım Şubat 2026
18 Takip Edilen4 Takipçiler
@lennysan This is true. I personally cannot do more than 2 for complex tasks. The context switching kills ya. I have heard people having one main agent and they talk to that only and the main agent talks to the sub agents etc.
English
"Using coding agents well is taking every inch of my 25 years of experience as a software engineer, and it is mentally exhausting.
I can fire up four agents in parallel and have them work on four different problems, and by 11am I am wiped out for the day.
There is a limit on human cognition. Even if you're not reviewing everything they're doing, how much you can hold in your head at one time. There's a sort of personal skill that we have to learn, which is finding our new limits. What is a responsible way for us to not burn out, and for us to use the time that we have?" @simonw
Lenny Rachitsky@lennysan
"Using coding agents well is taking every inch of my 25 years of experience as a software engineer." Simon Willison (@simonw) is one of the most prolific independent software engineers and most trusted voices on how AI is changing the craft of building software. He co-created Django, coined the term "prompt injection," and popularized the terms "agentic engineering" and "AI slop." In our in-depth conversation, we discuss: 🔸 Why November 2025 was an inflection point 🔸 The "dark factory" pattern 🔸 Why mid-career engineers (not juniors) are the most at risk right now 🔸 Three agentic engineering patterns he uses daily: red/green TDD, thin templates, hoarding 🔸 Why he writes 95% of his code from his phone while walking the dog 🔸 Why he thinks we're headed for an AI Challenger disaster 🔸 How a pelican riding a bicycle became the unofficial benchmark for AI model quality Listen now 👇 youtu.be/wc8FBhQtdsA
English
@zarazhangrui Nice. How many max agents do you use? Is it easy because you just talk to one agent? What about reviews and validation?
English
My ideal workflow:
1. In the morning, I have a 15-min standup meeting with a master AI who gives me progress of all subagents & tasks. I provide feedback & give instructions in one focused meeting
2. During the day, we all go our separate ways. Agents focus on their work; I focus on my work
3. In the afternoon, repeat step 1
Zara Zhang@zarazhangrui
My single biggest pain point right now: AI-induced attention deficit I have 5 Claude Code sessions running 10 Terminal tabs open 50 browser tabs open 100 X articles in my bookmarks When there are multiple AIs working for you, you're constantly task switching When you're waiting for AI to output, you're constantly distracted Because what used to be hard is now so easy, you always wanna do more, more, more Because of the speed of the AI news cycle, you feel a constant need to be on this app Add all this up and the result: you live in a constant state of distraction & delirium. Focus & deep work is harder than ever If anyone has figured this out pls teach me how
English
Markdown is not the answer to your AI problems.
It just can't be.
We need to force AI through deterministic gates.
English
@reactive_dude This should be good enough if you don’t have usage limit issues. The main issue is usage limits if you are not on the max plan. Codex / gpt 5.4 is good as well but you have to spend additional money.
English
Prash Krishna retweetledi
if you care about coding agents and tasteful software def go watch this talk by @badlogicgames it’s very good youtu.be/Dli5slNaJu0?si…
YouTube
English
@housecor This is insightful Cory. I also delibrate how lot pf the people boast of running parallel agents without sacrificing code quality. I can’t keep up with more than 3 agents and that’s hard with context switching.
English
I see many people running multiple agents simultaneously.
I avoid it.
Instead, here’s what I typically do while my *one* agent is running:
- Review the code it generates in real time
- Watch its terminal output and course correct
- Do code reviews for teammates
- Reply to email / Slack / Teams
- Plan my next steps
- Pair with a dev on a different task
- Pee
So, for me, running a second parallel agent is a last resort.
English
@nbaschez I haven’t solved it but I have been experimenting with reviewing once, implement one feature, test / eval and then iterate through planning again for the next feature.
English
My biggest challenge with vibe coding / agentic engineering lately has been getting stuck in what I call a "plan doom loop"
- have AI write a plan
- review myself, seems good
- have AI review plan, it always finds something
- repeat
It drains my time and energy to determine how important the "findings" really are
Who has solved this
English
@levie This is correct Aaron. You have to use it for
different use cases and get obsessed with it to actually understand what it takes to have it working for you at scale. You are either fully into it or out and there is no middle ground.
English
@Gregorein Yeah. He is probably trying to show anyone can be a tech entrepreneur or try a startup though there could have been a disclaimer it’s vibe coded and the reason for marketing it. It’s a fake it till you make it world.
English
pretty much, yeah. the discord-style shipping streak made me realy curious what his code actually looks like in production.
the audit was just a boring browser‑side curl of his own homepage + har taken from browser networking tools.
to make things worse i actually asked claude to do an audit too. a 2 sentence prompt.
English
so... I audited Garry's website after he bragged about 37K LOC/day and a 72-day shipping streak.
here's what 78,400 lines of AI slop code actually looks like in production.
a single homepage load of garryslist.org downloads 6.42 MB across 169 requests.
for a newsletter-blog-thingy.
1/9🧵
Garry Tan@garrytan
Absolutely insane week for agentic engineering 37K LOC per day across 5 projects Still speeding up
English
@romaindewolff An honest question which I am debating is do we
need such project mgmt software any more or can we build our own based on what’s needed for that person / company? The agents are good at listing and progressing thru their planned tasks or even a file based system works?
English
@DavidKPiano Exactly. Even if the users dont care about the code quality they want the app to work seamlessly when they want it and code quality is necessary for that to happen.
English
These takes are always wrong. Sure, users don't care about code quality *directly*
BUT they care about features shipping fast, bugs fixed quickly, & the app not degrading. That's all downstream of code quality.
A messy codebase is velocity-killing tech debt & users feel it.
Aaron Ware@EarnWhere
@thekitze Who gives a fuck about the quality of the code? Does it provide value? “Omg, I’m so glad I’m paying a subscription to this tool, the code base is soooooo clean and perfect” Said no consumer ever.
English
@jack @roelofbotha Very cool to see the move from middle
management to builders and owners. As mentioned in the article, it will probably break before it works but it would be good to stick with it.
English
our lead independent director @roelofbotha and i wrote about the history of organizational structures, and our intent to rebuild block as a mini-AGI. x.com/jack/status/20…
jack@jack
English
Prash Krishna retweetledi
Prash Krishna retweetledi
I built a new plugin! You can now trigger Codex from Claude Code!
Use the Codex plugin for Claude Code to delegate tasks to Codex or have Codex review your changes using your ChatGPT subscription.
Start by installing the plugin:
github.com/openai/codex-p…
English
@rohanvarma I think this is left to parents and not necessarily schools. Public schools will have issues getting AI fully adopted due to the concern that it may take teacher jobs away which is understandable.
English
I’ve been thinking about this a lot lately.
Imagine if a student today came out of high school not knowing how to use Google.
Fluency with coding agents like Codex is an asymmetrical advantage today, and hopefully a ubiquitous capability tomorrow.
I’m personally very passionate about ensuring we help schools evolve here. However, if you are a parent, I wouldn’t wait for the systematic change required here.
Jack Altman@jaltma
I hope schools are teaching kids to just sit down with codex / claude code and make stuff.
English
@noahzweben Noah this is good but due to the session limitis etc claude code is unusable for a long period of time. Reliability and stability should be the focus in my opinion.
English
Dispatch just filed 3 months of overdue work credit card receipts for me and that is AGI
English
@dansemperepico Because the larger portion of the world is non technical? Even if its a marketing wrapper removing the “code” should bring in the non technical people as it removes the scary part of the term code.
English
I don’t understand why Claude Cowork exists other than as a marketing wrapper to get non-technical people using agentic AI that we’re scared of trying Claude Code because they thought you have to be technical.
There’s nothing I can think of that Cowork can do that Claude Code can’t and I’ve had at least one experience where Cowork failed to do a non-coding task that Claude Code could do
English
@catalinmpit Interesting, any complex coding just burns the tokens quickly. I see my context jump from 0 to 50 pretty quick. Are you running Opus?
English
