dsivakumar's profile picture. Conversational Shopping Search at https://www.Tonita.co

Earlier:
Member, @southpkcommons;
NLP, ML, Algorithms @GoogleAI;
CS  Theory at @IBMResearch Almaden

.

D. Sivakumar

@dsivakumar

Conversational Shopping Search at https://www.Tonita.co Earlier: Member, @southpkcommons; NLP, ML, Algorithms @GoogleAI; CS Theory at @IBMResearch Almaden .

Pinned

Many AI and UI updates at Tonita.co, your personal shopper! Last month, we launched a beautiful UI re-design that is receiving a lot of love for its aesthetic and functionality. Last week, we launched a "virtual try-on" feature that has been delighting users!


with respect, while this is a consequence of G's dominant market position, they never told sites not to allow other crawlers -- AI or not. I run an AI shopping search engine that sends traffic to sites, and we're constantly blocked or throttled (made worse by Cloudflare's…

What’s Google Gemini’s unique advantage? It’s not the silicon. It’s that Google has access to more data because it’s tied AI crawl to search crawl. The Googlebot can go places online no other AI crawler can. 1/3

eastdakota's tweet image. What’s Google Gemini’s unique advantage? It’s not the silicon. It’s that Google has access to more data because it’s tied AI crawl to search crawl. The Googlebot can go places online no other AI crawler can. 1/3


The "Bracket City" game in @TheAtlantic is not only delightful to play every day -- it serves as a surprisingly cool "ARC-AGI" type problem to test frontier models' reasoning performance! Both GPT-5.1 and Gemini 3.0 Pro come close to solving these puzzles well with good…


D. Sivakumar reposted

The article also slanders non-commutative algebra.

boazbaraktcs's tweet image. The article also slanders non-commutative algebra.

Many AI and UI updates at Tonita.co, your personal shopper! Last month, we launched a beautiful UI re-design that is receiving a lot of love for its aesthetic and functionality. Last week, we launched a "virtual try-on" feature that has been delighting users!


D. Sivakumar reposted

asked Sora to give me a song where this was the hook; now it's on repeat and I really want a full version. some day

all i ship to prod is floats



congrats, although we've had this for several months now on Tonita.co :) and google's AI Mode seems to return an order of magnitude fewer results than we do (despite the fact that our corpus is a an order of magnitude smaller)

An exciting update today: you can now find, explore, and shop visual inspiration, right in AI Mode. Just describe what you’re looking for, upload a picture (or both!) and you’ll see a range of rich visual results that you can conversationally refine with follow-ups, instead of…



This paragraph from this article by @trishlaostwal is super-interesting. The methodology might have flaws, and users != usage (e.g., coding workloads on desktop), but the trend is interesting. Reminds me of 2012: mobile traffic (web, search) crossed desktop traffic. .

dsivakumar's tweet image. This paragraph from this article by @trishlaostwal is super-interesting.

The methodology might have flaws, and users != usage (e.g., coding workloads on desktop), but the trend is interesting.

Reminds me of 2012: mobile traffic (web, search) crossed desktop traffic.
.

NEW: Copilot is outpacing rivals on mobile. From March–June it added 5.6M U.S. users vs. ChatGPT’s 3.9M, per Comscore. Enterprise tie-ins and distribution are fueling Copilot’s surge, though ChatGPT still leads with 25.4M users. adweek.it/46kJHWM



Beyond "sparks" now... very nice, @SebastienBubeck ! side question: is it able to get the v2 result as well, with a bit of nudging / "brainstorming" in the right direction?

Claim: gpt-5-pro can prove new interesting mathematics. Proof: I took a convex optimization paper with a clean open problem in it and asked gpt-5-pro to work on it. It proved a better bound than what is in the paper, and I checked the proof it's correct. Details below.

SebastienBubeck's tweet image. Claim: gpt-5-pro can prove new interesting mathematics.

Proof: I took a convex optimization paper with a clean open problem in it and asked gpt-5-pro to work on it. It proved a better bound than what is in the paper, and I checked the proof it's correct.

Details below.


D. Sivakumar reposted

The challenge AI extends to those of us who write is to surprise ourselves, to travel far out of ourselves and to evade all mimicry by doing what we have never done before.


normally, we’d be #1 at this stage of the season… Google’s AI projecting from recent trend?? I’ve heard that ⁦@demishassabis⁩ is a Liverpool fan…

dsivakumar's tweet image. normally, we’d be #1 at this stage of the season…

Google’s AI projecting from recent trend??  I’ve heard that ⁦@demishassabis⁩ is a Liverpool fan…

we want to hear your "that's another stor[ies]", Eli

How did I end up a Mathematician? (long story) As a child, my dad said to me "our family isn't good in math" and I, preferring soccer and dating to anything involving brain work, accepted this as truth. As an undergrad I studied Biology and CS, to go into Neuroscience (I ended…



The new strawberry... 🍓

dsivakumar's tweet image. The new strawberry... 🍓

how hard is it to add a line to a prompt telling the LLM to write like a very smart, sharp, near-genius who's slightly flawed in terms of spelling and grammar??

How long before people intentionally put in typos to make sure people believe an LLM didn’t write a thing?



Loading...

Something went wrong.


Something went wrong.