Annanth Aravinthan @ProtagorasTO

Previously CRO at EliseAI and Head of Business Ops (Launch, Strategy, Finance and Competition) at Uber Eats Canada annanth.substack.com/p/4-food-deliv… Joined November 2013

Tweets

4K
Followers

8K
Following

298
Likes

9K

Annanth Aravinthan @ProtagorasTO

4 days ago

@thaAdamLittle In the US, TCPA prevents automated "robocalls". I don't think that'll change. So, SDRs will transition to largely outbound calling with other tasks being centralized and automated.

0 0 0 536 0

View Details

Annanth Aravinthan @ProtagorasTO

6 days ago

@marketeuphoria @devahaz AI has strong PMF in images and design. So even a subpar product is going to drive incremental revenue growth. But it also creates new white space for start-ups to get to scale, which allows them to attack Adobe's core customer base. This can take a while, like 5-10 years.

0 0 1 20 0

View Details

Annanth Aravinthan @ProtagorasTO

a week ago

@apoorv03 Probably aligns with the incremental value people are getting out of the tokens?

0 0 1 639 0

View Details

Annanth Aravinthan @ProtagorasTO

a week ago

OAI rate limit resets remind me of Uber promotions in 2012. "Refer a friend and get $20 off your next ride".

Greg Brockman @gdb

a week ago

For next two weeks, refer your friends to Codex, and you'll bank a rate limit reset:

131 80 1K 170K 146

0 1 0 670 0

View Details

Andrew Tan @andrewztan

a week ago

I'm hosting a pickleball / padel event in SF on Wed 6/17 with @elise_ai and @SapphireVC for AI engineers, builders, and founders! Come by for some friendly competition and good vibes 🎾 luma.com/l0cu9a2z

1 7 22 981 7

View Details

Annanth Aravinthan @ProtagorasTO

a week ago

Thanks! To clarify, to me frontier = you have the best models *now* (so there's a timing component to it). Which is different from they need very capable and powerful AI *eventually*. I think the latter is true. There's no way to avoid the training (and research) costs. FYI a lot of the capex is actually going towards future inference to serve these models.

1 0 1 51 0

View Details

Annanth Aravinthan @ProtagorasTO

a week ago

@KairosPraxis Re: $META.. 1. Below I believe is a summary of Facebook's medium term opportunities with AI 2. I actually don't think they need to be on the frontier to make make most of this a reality 3. It seems like they are "catching up" quickly and people will be surprised by end of 2026

2 1 3 1K 4

View Details

Annanth Aravinthan @ProtagorasTO

a week ago

@Noahpinion All of this is downstream of energy.

0 0 0 725 0

View Details

Annanth Aravinthan @ProtagorasTO

2 weeks ago

@nachkari I'm not sure if it shows up in GDP, that's sort of my point. Property manager materially reduces Opex in labour / wages, most of that passed as consumer surplus, some kept as operating margin.

1 0 0 33 0

View Details

Annanth Aravinthan @ProtagorasTO

2 weeks ago

@nachkari Also check out the below example.. value creation is real.. unsure of how this shows up in GDP and adoption takes time. x.com/ProtagorasTO/s…

Annanth Aravinthan @ProtagorasTO

7 months ago

Strongly disagree with @dwarkesh_sp on this one. Since o3, models can automate *most* white-collar work. I'll give you an example of a real life deployment of an AI agent: >$1M ARR contract - One year sales cycle just to get a pilot started - Had to hire an entire team (exec +

0 1 4 2K 3

1 0 0 35 0

View Details

Annanth Aravinthan @ProtagorasTO

2 weeks ago

GDP isn't perfect (doesn't measure changes in price or quality well). ChatGPT use cases are mostly editing and helpful info (finance, health, etc). Mostly about convenience. Trivial impact to GDP. Anthropic I think is way temp over-earning vs. incremental value creation: x.com/ProtagorasTO/s…

Annanth Aravinthan @ProtagorasTO

a month ago

The Anthrophic Krishna Rao with @InvestLikeBest is great. Says they want to be a platform. Take Gates postulate that platforms capture small % of value for customers / ecosystem. iPhone - $200B rev = $2T in value Windows - $17B rev = $170B in value Anthrophic - $45B rev = $450B

0 0 2 273 1

2 0 0 68 0

View Details

Annanth Aravinthan @ProtagorasTO

2 weeks ago

Sarah Friar on the All-in podcast: "Our next big training run in the Fall will be on Vera Rubins"

0 0 0 182 1

View Details

Annanth Aravinthan @ProtagorasTO

2 weeks ago

@signulll You'd be surprised by how many things are not automated because of resource constraints. For example, for a long time, the restaurants you could see when you opened the app and we were short on drivers were manually modulated by a human being in the Philippines.

0 0 9 1K 0

View Details

Annanth Aravinthan @ProtagorasTO

3 weeks ago

@lpolovets I think this is like using the taxi TAM to determine Uber's TAM. Coding's TAM would encompass automating *most* business processes for knowledge workers. So likely trillions of dollars. Of course, most of the future things to automate haven't been imagined yet.

0 0 3 507 0

View Details

Annanth Aravinthan @ProtagorasTO

3 weeks ago

@nachkari Not an expert but auto an important industry given dual use with military production?

0 0 0 101 0

View Details

Annanth Aravinthan @ProtagorasTO

3 weeks ago

the answer is.. very long (2-3 years).

0 0 1 285 0

View Details

Gavin Baker @GavinSBaker

3 weeks ago

Composer 2.5 being Pareto dominant in coding per CursorBench is important. This is after only a few weeks of supplemental training and/or RL in the Colossus 2 cluster. The 1.5 trillion parameter version of Grok will likely be a much better base model than Kimi. We shall see.