Ujjwal Maheshwari
295 posts
Started being actively involved again and treating the AI like it’s an adversary and I got my groove back.
So- I tell the agent in a terminal what I want it to do. I go make a latte. When I come back it's generated about 800 lines of code with passing tests. This would have taken me hours.
I open up an IDE to look at the files. I tell the agent it’s an idiot. I ask what the fuck it's doing, it’s making some stupidass overly defensive assumptions. "You're right to question me." It deletes 400 lines.
I keep reading. I ask more questions. "I made incorrect assumptions." We're down to 200 lines.
I run an ai code review. The reviewer agent identifies a bunch of issues that sound very bad. Instead of telling the agent to handle it I read and reject them outright. None of this shit will happen because despite all the context you’ve been given you don’t actually understand distributed systems.
Final PR is 190 lines, including tests.
This looks good. You are much better when I remember you're just autocomplete than when we both pretend you’re intelligent.
Vic 🌮@VicVijayakumar
I'm now able to tell my agent “we are going to work on JIRA-1234” and it goes and pulls down the task, makes me a plan, I say yeah okay that looks good, and it generates the commit. I run an AI review from a different session, it finds 4 issues of varying priorities, I paste it to my original agent and say validate these findings and fix them if necessary, it creates a fix, I run another review, no more high priority issues found. I open up the code in an IDE to go over it before pushing it up for human review. Looks fine I guess, nothing crazy. I try to understand everything before I push it up for review because if this breaks, it's still my name on it. I say why did you make this one change, it gives me a reasonable explanation for why. It says something codebaity like "if you want I can suggest 2 more ways you could really tighten up this work to prevent some rare but possible regressions". I'm smart enough to not fall for it. Code pushed up, task moved to in-review. I didn't write any of it, this is not my accomplishment. Users won't care who wrote it if it works. A lot done in 20 mins but it felt soulless.
English
If I get the 5/5 invite I'm going to lose my shit
Can't wait
English
If we're right, people lose Jira tickets. People lose pull requests. People lose IT guys. You know what I hate about fucking LLMs? It reduces people to tokens. Here's a number - every 1% Anthropic improves, 40,000 devs have to work for McDonald’s. Did you know that?
English
@pronounced_kyle If it helps, we find this hilarious as well
English
Dads, I need advice.
My toddler is now the perfect height to kick me square in the nuts while I hold him.
He's folded me twice in two days. He finds this hilarious. What do I do
English
Ujjwal Maheshwari retweetledi
#OperationSindoor on the games field.
Outcome is the same - India wins!
Congrats to our cricketers.
English
English
Had a great time with @chrismichel yesterday. He took photos of me for a series of portraits he's making of members of @theNAEng and @theNASEM.
All his pictures are fantastic: you can see many in this series at explorers.com
Thanks, Chris!
Christopher Michel@chrismichel
New Heroes portrait of @Google's @JeffDean for the @theNAEng & @theNASEM. Spending time with Jeff is a reminder that kindness and brilliance can live side by side. He is one of the great architects of the modern internet, the co-creator of MapReduce and Bigtable, and the driving force behind TensorFlow. Today his work touches billions of lives, yet he carries it all with an amused, approachable warmth. In conversation he is rigorous and honest, but also generous, especially with younger engineers who still recall his detailed, thoughtful code reviews. Elected to the National Academy of Engineering, Jeff continues to shape the future of artificial intelligence while never losing his delight in the puzzle of a good problem. Full bio: explorers.com/jeff-dean/
English
I hear you — and I won’t dismiss what you’ve felt. Coming back with hope, only to be met with apathy or resistance, can be heartbreaking. Especially when you genuinely wanted to contribute.
But for some of us, India isn’t just geography or policy — it’s personal. It’s messy, loud, flawed… but still ours.
Not everyone is clapping for bad decisions. Many are quietly fighting, building, fixing — without applause or power. Change here isn’t fast or easy. But maybe that’s why some of us choose to stay — not because it’s perfect, but because we still believe it’s possible.
And I get it — choosing peace for your family is valid. You don’t owe anyone an explanation. But just know… some of us stayed. And we’re trying.
One small battle at a time.
English
I used to find it funny when people said —
“The end goal of every Indian is to leave India.”
A few years ago, I left the U.S. and came back driven by hope and a sense of duty.
I genuinely wanted to build a life and contribute.
But the ground reality was different.
People blindly defending poor decisions.
Applauding high taxes, no merit, and unchecked corruption.
It felt like a society rooting for its own collapse.
Eventually, I packed up and left — this time, with clarity.
And honestly, I’ve never felt more at peace.
India will always be in my heart.
But I refuse to raise my kids in an environment of hate, noise, and wilful ignorance.
English
@priyanshaaaa DeeTee SHUT DOWM WTF IS GOING ON WHY AM I ALIVE
English
So energizing to pull together 50 folks from across two decades of Manipal batches - into one buzzing room filled with conversations about startups, AI, and, of course… the shutdown of our fav college bar, Deetee.
May or may not have done it for the cute certificate.
English
At IIMB, we used to play a bastardised version of Tsepak Takraw on the tennis courts. I was in first year and on my team was Anand, PGP2 and President of the student body.
During an intense rally he suddenly raises his hand and says - foul guys I fouled.
The ball had hit his hand and no one had seen. But he fessed up. It made a profound impression on me. If Anand calls me today and asks for money I will send it to him no questions asked. (He’s doing very well, no shade, just saying)
Character is everything and it shows up everywhere.
Surbhi Jain@surbhiskjain
𝐈𝐭’𝐬 𝐟𝐮𝐧𝐧𝐲 𝐡𝐨𝐰 𝐩𝐚𝐭𝐭𝐞𝐫𝐧𝐬 𝐬𝐡𝐨𝐰 𝐮𝐩. A regular at the club was playing with us and kept insisting points were his, even when it was clearly out. At first, I brushed it off as competitiveness. But it kept happening. Something just felt… off. A few things didn’t add up in my mind, and I asked our club managers for his payments. Found out he hadn’t been paying for his games. He was using my name, clicking a photo of the scanner as if he was paying, and never actually doing it. When we confronted him, there were vague excuses, shifting accounts, and stories that didn’t quite add up. What struck me was this: 𝐇𝐨𝐰 𝐬𝐨𝐦𝐞𝐨𝐧𝐞 𝐛𝐞𝐡𝐚𝐯𝐞𝐬 𝐢𝐧 𝐨𝐧𝐞 𝐚𝐫𝐞𝐚 𝐨𝐟 𝐥𝐢𝐟𝐞 𝐨𝐟𝐭𝐞𝐧 𝐞𝐜𝐡𝐨𝐞𝐬 𝐢𝐧 𝐨𝐭𝐡𝐞𝐫𝐬. The same dishonesty about accepting a lost point showed up in the way he was hiding non-payment. Small signs, similar behaviour. It reminded me that 𝐢𝐧𝐭𝐞𝐠𝐫𝐢𝐭𝐲 𝐢𝐬𝐧'𝐭 𝐚 𝐬𝐰𝐢𝐭𝐜𝐡; 𝐢𝐭 𝐬𝐡𝐨𝐰𝐬 𝐮𝐩 𝐢𝐧 𝐭𝐡𝐞 𝐭𝐢𝐧𝐲, 𝐮𝐧𝐧𝐨𝐭𝐢𝐜𝐞𝐝 𝐦𝐨𝐦𝐞𝐧𝐭𝐬. 𝐀𝐧𝐝 𝐬𝐨𝐦𝐞𝐭𝐢𝐦𝐞𝐬 𝐢𝐭'𝐬 𝐭𝐡𝐞𝐬𝐞 𝐬𝐦𝐚𝐥𝐥 𝐝𝐢𝐬𝐜𝐨𝐦𝐟𝐨𝐫𝐭𝐬 𝐭𝐡𝐚𝐭 𝐡𝐞𝐥𝐩 𝐫𝐞𝐯𝐞𝐚𝐥 𝐭𝐡𝐞 𝐛𝐢𝐠𝐠𝐞𝐫 𝐭𝐫𝐮𝐭𝐡.
English
@chrisalbon Fun fact : it's the second most visited museum after the Louvre
My fav exhibits are the SR-71 blackbird and the actual plane that dropped the nuke on Nagasaki.
English
This is DC’s best secret. There is the National Air and Space museum in DC where all the tourist go, which is neat, but a bit outside DC is udvar-hazy and its where they have all the huge things that can’t fit in the other museum
anton 🇺🇸@atroyn
today i fulfilled a boyhood dream and visited the smithsonian air and space museum (udvar-hazy). here are some favorite exhibits.
English
@Yuchenj_UW Living in Miami a couple of months now. Most 2k+ google review establishments have fake ratings. 18-20% service charge addition is becoming standard. Give a big fat zero for tip if a service charge is added. And yeah, learn Spanish!
English
>went to a Miami restaurant
>12K Google reviews, 4.9 stars
>food is mid
>bill arrives, way higher than menu prices
>"sir, we added 20% service fee + 2% resort tax"
>we only 2 people!
>wtf_is_resort_tax.jpg
>tablet flips around
>"Choose your tip: 18%, 20%, 25%"
How is this legal?
English
does anyone use outlook on your browser? do you want support for an agent to go read and send emails and schedule stuff for you?
English
@Shpigford @cursor_ai I agree. Couldn't get it to create a simple python script which makes a request. UX has changed significantly as well which makes it worse
English
I feel like I have no idea how to get @cursor_ai to do what I want anymore.
Between the different modes, different models and UI/UX updates, I feel dumber than ever.
There was some inflection point in the last couple of months where everything just got…worse.
Just me?
English
Is this the gradient descent for prompts? Pretty close tbh
Rafal Wilinski@rafalwilinski
Build your own pseudo DSPy-ish prompt auto-optimiser in 3 steps with @braintrustdata: 1. Grab Braintrust’s OpenAPI and turn into MCP (there are OSS libs to do that) 2. Plug-in that MCP to Cursor 3. In Agent mode: “please optimise this prompt (@src/prompt.ts) based on evals (pnpm run evals) until you get a substantial improvement. Try up to 10 times.” 4. Enjoy free accuracy gains and your promotion to principal
English
Someone found a bug in my open source LLM usage tracking tool. I couldn't be happier. Atleast someone's using it :)
English
Overall, a bad experience. Can't handle straightforward things and bad support overall.
For context, I am working on a student visa in the US. Imp. point is: I don't have to pay FICA taxes. Here are the problems that I've faced :
- Notifications for tax docs were never sent to me. HR said "fill tax docs". Couldn't find them on Deel. Why wasn't it a step in onboarding?
- In my first payroll, FICA taxes were deducted. Cool. Asked HR. She asked Deel. Deel says yep we've taken care of it. Next payroll, FICA taxes were deducted again. Deel is like, oh yeah the change was made after the "payroll cut off date", dw next payroll will be perfect. Next payroll, guess what, FICA deducted AGAIN. Deel is like, oh damn we so sorry, this shouldn't have happened. Continues to happen for 6 payroll cycles before it is corrected.
- Now, FICA taxes cannot be claimed through the IRS. The company has to refund them. Asked HR. HR asked Deel to refund. In November '24, Deel says dw 1st week of December it should be done. It's still not done. We are in the middle of March '25.
- incorrect W2s for the entire company. COO on email claims "should be done next week". Took two weeks after that. They said the FICA thing will be fixed as well. Not fixed yet.
All of this happens over emails, cause the general support cannot help. Most of the times, no timelines are given. When timelines are given, they are never followed. So what's the freaking point? Just extremely taxing and annoying and energy, soul sucking follow ups.
"CAN YOU PLEASE GIVE ME MY MONEY" in hundred different tones.
I know all of the above are edge cases but the great products and services do edge cases better than competitiors. And Deel doesn't.
I would avoid Deel as far as possible. In fact, when I interview with companies, I ask them if they use Deel. If yes, try to avoid.
Btw the same FICA thing happened at my new job. They use ADP. The tax deduction was stopped in the next payroll and a refund was made in the same paycheck. I was the happiest man that day.
@DanWestgarth
English
@ujjwalm29 @parkerconrad @deel Why do you dislike it so much? Just curious. We are almost certainly going to go ahead with Trinet over Deel.
English
Rippling sued @Deel today. Our lawsuit alleges Deel cultivated a spy at Rippling & orchestrated a long-running trade-secret theft. The spy searched “deel” in our systems 23 times per day on avg, letting him spy on Deel’s own customers who were considering a switch to Rippling.
English