Sagar Hebbale

Guilty. This is exactly how i got into software.. In college I did my grad course in electronics/hardware. Most jobs that came through campus were software roles, never wanted to interview for any of them. Then I heard about an opening developing mobile games. iOS developer. Suddenly my lack of interest in software didn't matter. I didn't even know how to code. Got the job on pure interest alone. 16 years later, I'm running a software company.

every software engineer is just someone who wanted to make video games and quietly gave up on that dream

@KaiXCreator Both! But this is changing, as pure models go my preferred model for coding is slowly becoming gpt 5.5. Used to use opus 4.7 a lot before

Are you team Claude or Codex?

Tried to single shot a long running task with claude opus 4.7 extra high. Did not work and burnt 60$! 😢

@piersmorgan Its clear who deserves this award

WHAT IS GOING ON? How can a serial loser this season beat a man who has driven Arsenal to our first League title in 22 years? This is complete madness.

@unclebobmartin Just a matter of time till we get an actual star trek pin/badge wearable that actually gains mass adoption! Wonder if the LLM companies will get into hardware or if its a company like apple that will do this

These agents are the Star Trek computers. You can ask them question and get answers just like Spock and Kirk used to do. It's uncanny.

Dont ask LLM’s to do math.

@ManCity What a legend. City will never be the same again. Thats how special this time was!

"What a time we have had together." 🩵

@GuptaSanskritii Don't one shot it. Split up the work into smaller tasks and ask it to execute. You also have to monitor and provide feedback on the smaller tasks...it compounds to what you want

Dude claude is not able to make one single onboarding screen according to my figma design. I dunno what am i even doing wrong atp.

@gdb Someone once also said "LLM's will not lead to AGI" @ylecun

the model alone is no longer the product

@sama Help us crack advanced (non rocket) propulsion systems

what problem do you most hope AI will solve in the future? maybe we can help!

@icanvardar Not if you maintain a TECH-DEBT.md and keep updating it 😜

vibe coding is just speedrunning technical debt with confidence

A small story about why I won't trust @porterit_ with anything valuable again. Booked a Porter to send a friend some organic farm-grown mangoes and homemade pickles. Rider arrived, collected the package, and then was never to be seen, heard, or contacted again. In his defense: peak mango season, the mangoes smelt phenomenal, there were pickles. A weaker man would've also disappeared. But my calls went unanswered. And Porter support's grand strategy was to keep closing my complaint, presumably hoping I'd disappear too. Luckily nothing valuable was lost. Just my faith in Porter for anything that actually matters.

@carllerche The programming language that the LLM was most used in training data maybe? So javascript? 🤷🏽

Which programming languages will thrive the most in the agent-era? Why will it not be Go? It is simple, consistent, predictable, and has excellent local reasoning.

@theliverpoolway Are there @Polymarket odds for this?

Bet your mortgage on them getting a red card or conceding a penalty this weekend now that it doesn’t matter. Nailed on.

@oprydai And here are are, trying to make computers human

humans are biological computers

@DanKulkov Ive been doing this a lot of late! And it is fun

fuck marketing building fun project launch tomorrow

@rand_longevity How do you see yourself?

what is the first question you will ask the AGI?

The principle should be: more work, fewer people, not more people, less work. Agree the nature of work is changing, but I'd push back on the "more layers" point. If AI genuinely makes businesses more competitive, it shifts the optimal org size downward in a lot of sectors, and large monolithic organizations become the most exposed, precisely because their inertia is the liability. The future looks more like small teams innovating at the scale of today's large companies, and more human decision layers cuts against that.

I’m not a doomer an AI at all. I think the nature of work, particularly entry level jobs will change. Ai will make business more complicated and competitive. Not less. Which means more layers where humans have to make decisions before the next process can happen And there will be millions, if not more, local models. We will we modify and train them for our lives and the businesses we create. Just leveraging the foundational models for things we don’t have access to. Maybe the new gig economy will be the models that represent our lives and the knowledge and experiences each one of us train those models on, and we get hired by companies for access to those models ? Thoughts ?

GPT 5.5 is my go to LLM for coding these days. But it wasn’t always. How has your choice of LLMs for coding evolved over the years?

@OpenAI We have come a long way in just a few years from Wolfram alpha

Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946. For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids. An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better. This marks the first time AI has autonomously solved a prominent open problem central to a field of mathematics.