Arvind Goyal

Senior Engineering Manager - AI @gohighlevel previously built and led engineering at @getconvinai | @shopflo_ | @unitedhealthgrp | @tinder | NIT Jalandhar

[Arvind Goyal @Imarv7]

while openai, anthropic build models to kill engineers, I feel software is going to be extinct soon so there would be no use case of an app.

We needed code and APIs to communicate when we had UI apps.

Every app becomes an agent. Who will build an app?

[Arvind Goyal @Imarv7]

people working in tech/ai - how are you feeling?

[Arvind Goyal @Imarv7]

@heeriamok Can’t find the girls who don’t smoke.

[Arvind Goyal @Imarv7]

@harshilmathur I think it’s solvable - if the new products can come from anywhere in the hierarchy. To solve this - have transitioned AI teams to Team of One concept/FSB. Everyone is a builder, designer/PM/engineer/QA - ideas/prototypes can come from anyone, with a loom - not just a leader.

[Arvind Goyal @Imarv7]

@AshwiniVaishnaw @INCIndia @kharge 140 crore - 1. I want to be insulted like this 24/7.

[Arvind Goyal @Imarv7]

“we” who we? Is it the @BJP4India party’s own funds?
Or is it from “we” the taxpayers of India, solemnly swear to bribe non-taxpayers to vote for a morally corrupt government?

[BJP @BJP4India]

All unemployed youth in Bengal will receive ₹3,000 in their bank accounts on the first of every month.

We will provide ₹21,000 to every pregnant mother for the care of her child.

The BJP government will implement 33% reservation in government jobs, and from 1 June, all women will get free bus travel.

The BJP government will increase the PM-Kisan Samman Nidhi from ₹6,000 to ₹9,000.

- Shri @AmitShah

#BanglarMoneSudhuiBJP

[Arvind Goyal @Imarv7]

@AmitShah Why link it with delimiting?

[Arvind Goyal @Imarv7]

Feel like going for master’s in physics now that software is done

[Arvind Goyal @Imarv7]

Never write off zuck.

[Alexandr Wang @alexandr_wang]

1/ today we're releasing muse spark, the first model from MSL. nine months ago we rebuilt our ai stack from scratch. new infrastructure, new architecture, new data pipelines. muse spark is the result of that work, and now it powers meta ai. 🧵

[Arvind Goyal @Imarv7]

@ishaaniguess The agent TTS the “thinking” to me(addressing me in third person), it also is going through the checklist and speaking it out loud to me if it should ask me something or not- thinking should be kept in background, I think just give it a shot and you’d know what I mean.

[Arvind Goyal @Imarv7]

@NirantK Have similar kind of experiences working for startups, learning was great but ESOPs fell short of dreams sold, no hate to the founders, they’re good folks but the equity promises feel hollow when we have such experiences - especially when employees risk/invest by joining early.

[Arvind Goyal @Imarv7]

Saw my hair get thinner for a year or two, didn’t bother to see a doctor until it got bad, only to find out I have severely low levels of Vitamin D and B12 - which isn’t effecting just hair health but lot more.

If you’re a vegetarian, this is a sign to get your vitamin tests.

[Arvind Goyal @Imarv7]

@vinitsarode_ 27 yo should be one Mike Ross

[Arvind Goyal @Imarv7]

@contextkingceo It was funny until you put an ad on the twt and now it’s not

[Arvind Goyal @Imarv7]

They don’t know what they’re doing - memory is just a .md file, you’re better off to an obsidian kind of arch, an SDE1 in your team will be able to do memory - you can ignore such people and their products.

We CALLED OUT THE BULLSHIT, and they’re doing damage control.

[Dhravya Shah @DhravyaShah]

We scored 99% on LongMemEval to prove a point.

Eval results aren't the only thing that matter.

[Arvind Goyal @Imarv7]

Some people on twitter, and they’re in double digits, make low IQ engagement farming tweets, it’s similar to the last bencher in classroom who just wants to grab attention by making others laugh, all this for few $, this at at cost of their credibility on anything meaningful.

[Dustin @r0ck3t23]

Sam Altman just told a room full of students where the real opening on the board is.

The global market assumes the Transformer architecture is the final form of the compute engine.

It’s a stepping stone.

Altman: “I bet there is another new architecture to find that is gonna be like as big of a gain as transformers were over LSTMs. And I think you finally have models that are smart enough to help do that kind of research.”

The frontier models aren’t replacing human researchers.

They’re accelerating them.

The operator who uses today’s AI to hunt for the successor to the Transformer doesn’t just win the current hardware cycle.

They lock in the next one before anyone else knows it exists.

And here’s what that means for products.

Altman: “I would just pick a big area and say what is possible now with AI that wasn’t possible at all before. Like where can I totally redo something that’s like, AI is the absolute core to the interaction working.”

The traditional enterprise is trying to bolt an AI chatbot onto their existing software suite to appease shareholders.

Winning operators are burning the entire product category to the ground and starting over from zero.

You don’t optimize the friction of an archaic system.

You build an entirely new architecture where artificial intelligence is the foundational bedrock.

The builders capturing the board over the next decade won’t build “AI-assisted” tools.

They’ll look at massive, entrenched industries and realize the entire category can be rebuilt from scratch by a single, AI-native execution loop.

But here’s where it gets exponential.

Altman: “AGI will look like just a warmup for whatever the next important thing was. And that’ll keep going for the rest of history.”

AGI is not the destination.

It’s the warmup lap.

Altman told a room full of students this is “at least the best time ever so far.”

Not because the race is almost over.

Because it’s barely even started.

Every breakthrough from here compounds into the next one.

The organizations treating this as a gradual transition are training for a finish line that doesn’t exist.

[Arvind Goyal @Imarv7]

Opus 4.6 have been hallucinating and failing so many times for me, Codex 5.3 is way more consistent and faster.

Not sure what goes on at times when we just spend so many tokens and it’s all garbage - every token costs equal today, even it llm hallucinates which is kinda crazy.