Yash Shanker

Engineer

[Yash Shanker @DevOoopsEngnr]

@YashShanker Yash, i hope you got the help needed.

I need to speak with you regarding something else, could you open your DM please, or message me in DM.

Thanks.

[Yash Shanker @DevOoopsEngnr]

This has become more common than people would tend to believe. I recently had to build a test harness, whose only job is to scout for compute capacity across 15 zones across regions across market-options and launch compute to run tests in the first shot.

[Richard Crowley @rcrowley]

Honest to god, I got an InsufficientInstanceCapacity error for a t4g.nano in a test suite today. The cloud is FULL.

[Mitchell Hashimoto @mitchellh]

I've got an agent in a loop optimizing a renderer with the goal to minimize frame times (and tests to measure). It got times down from 88ms to 2ms and allocations down from ~150K to 500. Sounds good, right? Wrong. This is exactly why agent psychosis is a big fucking problem.

As an experiment, I rewrote the Ghostty core render state in Go, with access to identically laid out data structures as Ghostty and the exact same validation tests. I made a purposely naive renderer (simple, correct, but slow). 88ms per frame with 150,000 allocations (horrendous, lol)!

I then kickstarted a Ralph loop to bring the frame times down. I told it it can't modify input data structures or the public API or tests (they're correct), but it can do anything else it wants. It got to work.

It has worked for about 4 hours. I've spent around $350 on this experiment so far. The results?

88ms => 1.5ms
150K allocs => ~500 allocs

Incredible right? Nope.

My hand-written renderer I ported has frame times (same benchmark) of ~20us (0.020ms) and 0 allocations in the update path.

This is the problem with psychosis and lacking systems understanding. If you don't understand the system, you're going to accept that this is an incredible result. If you understand the system, you'll see better solutions immediately and can do roughly 75x better on throughput.

The people who blindly trust agent output are in the former camp. They're sheeple, overdrinking from a fountain of mediocrity.

Standard disclaimer: I use AI all the time. I like AI. The point I'm making is to not blindly accept results. Think. Analyze. Learn.

[Arpit Bhayani @arpit_bhayani]

Surround yourself with people whose default behavior is your desired behaviour.

[Yash Shanker @DevOoopsEngnr]

@daniel_mac8 @openclaw @thdxr i guess this doesn’t apply to us folks who used opencode with claude models with claude pro/team/enterprise plan, and not with API key ? Or does it ?

[Mutha Nagavamsi⚡️ @MuthaNagavamsi]

never stop learning kubernetes. because there is a huge upside. but...

always use kubernetes only when it's necessary. I repeat, use only when it's absolutely necessary

[Uncle Bob Martin @unclebobmartin]

A team who allows messes to persist in their code will gradually slow down to a tiny fraction of their potential productivity. I’ve seen teams that have expanded their estimates out to weeks and months for jobs that ought to take a day or two.
The slower the team goes, the more the pressure builds. The more the pressure build the more messes are left uncleaned. Each such mess impedes the team even more; and the slower they all go. It’s a vicious cycle, and if you’ve been a programmer for more than a couple of years you’ve likely taken a few turns around it.
But it gets worse. Managers, desperate to increase productivity, hire new people. This, of course, has the opposite of the desired effect. The new people, thrown into the mass of messy code, see how things are done around here and — of course — they emulate it. So now there are even more people making messes, which drives the cycle even faster, and drags the productivity even lower.

[Yash Shanker @DevOoopsEngnr]

@imJenal In.

[Yash Shanker @DevOoopsEngnr]

I’ve done this for the last 8 years and it’s a damn hack for getting work done.

So much so, that in my last org when my reports used to take random PTOs, I’d sometimes know we’re getting shit done.

[Jaana Dogan ヤナ ドガン @rakyll]

Faking PTO to write code is a globally useful tip.

[Yash Shanker @DevOoopsEngnr]

Take my money, don’t give me equity, just do what you’re saying you’ll do in the first statement.

[Y Combinator @ycombinator]

Operand (YC W25) is building an AI to kill McKinsey—starting in e-commerce & retail.

They’re already driving 6-figure P&L gains for brands. The future of consulting isn’t slides.

https://t.co/uumBYpHmfe

Congrats on the launch, @ak_iyengar, @realajsfr, and @r_gorthi!

[Yash Shanker @DevOoopsEngnr]

@martin_casado 🙋🏻‍♂️

[Yash Shanker @DevOoopsEngnr]

@prabodh_agarwal Uhu uhu
* coughs in yaml

[Yash Shanker @DevOoopsEngnr]

@Purnanga01 @mufaddal_vohra Nah bro, didn’t meet them, but they were really close to where I sat.

[Yash Shanker @DevOoopsEngnr]

💪🏼🚀

[Carl Pei @getpeid]

Bullish on India

[Yash Shanker @DevOoopsEngnr]

@ilah108 Vermeer

[Yash Shanker @DevOoopsEngnr]

What a story ❤️
Truly a 🚀

[Amir Salihefendić @amix3k]

Todoist has made more than $100 million in total revenue, which isn't very interesting because many others have reached this number. What's interesting is that we did it in our unique way:

— Fully bootstrapped. Customers have supported us since the beginning, and we've used only our revenues to improve things further.
— Complete independence. Since we are customer-supported, no one tells us what we can or can't do.
— Remote-first. But not only remote-first, we've hired super talented people worldwide who never went to Ivy League schools or worked at Google. Many of the early people who joined Doist have seen a 10x increase in compensation as we've scaled Doist.
— Europe mixed with US mentality. We've achieved this by working 40-hour workweeks and taking 40 days of vacation per year. We only work on weekdays. The three-member CXO team has 8 kids, and we got them while we scaled Doist.

The most critical personal lesson I've learned is that you can do much more than you think. I came from a refugee background and started a real school in the 4th grade. Starting and running a tech company was not even plausible while growing up. But here we are! So start learning, growing, and building! You got this 😊🚀

[Yash Shanker @DevOoopsEngnr]

@nikhiljoisr I slept the entirety of the first innings.
Been more of a fanboy for the bowling attack these days.

[Ethan Evans @EthanEvansVP]

I got to Amazon VP after 2 promotions by managing up well. My managers fought for my promotions.

You get "fight for you" manager support by building a partnership and meeting 3 criteria that get your manager to put their reputation and energy into helping you:

[David de Gea @D_DeGea]

Manchester is and always will be RED! 🔴🔴

[Chris @criccomini]

Object store all the things!