Siddharth

Parenting | AI × Product | Badminton | Fitness | Arsenal | Raja, Harris & U1 | ex-Amazon, IIM-A

[Siddharth @dash_third]

https://t.co/PJHTP01mXR

[Siddharth @dash_third]

End of season update.

The question I started with: Can a self-improving system learn to model T20 madness, or is it just noise?

With a 23-match sample, it's mostly noise. Model predictions were close to a coin flip. Interestingly, even Polymarket predictions (with real money riding on it) were close to a coin flip as well.

That said, I find it cool that the model taught itself to land right at that same line with no human in the loop. 91% of results fell inside its prediction bands.

Whether that is learning or just finding the floor, I dont know because sample is not big enough. Maybe next season.

[Siddharth @dash_third]

Fun research experiment I'm running through this IPL.

5 IPL matches in. Polymarket seems to overprice narratives (dew, form) and underprice boring venue-specific base rates (head-to-head at the ground, chase win rates, bowling matchups). Matches 4 and 5 especially.

Based on a multi-agent system that I built to predict match outcomes. Fully autonomous on a VPS, no human in the loop and self improving via Autoresearch (thanks @karpathy sensei).

n=5 so grain of salt. Repo is public 👇

[Siddharth @dash_third]

Repo link: https://t.co/8bKwaXKf0C

[Siddharth @dash_third]

This is cool. If Codex can generate good frontend, makes it a lot easier to switch from Claude code imo

[Webflow @webflow]

Excited to partner with @OpenAI Sites as we help bring the next generation of web creation to life.

[Siddharth @dash_third]

Core memory unlocked ✅ Being with 10 lakh arsenal fans (or probably more) singing and celebrating 😍

[Siddharth @dash_third]

@championswimmer "that nobody wants to talk about"

[Siddharth @dash_third]

This is a 🐐 video. Hilarious

[Siddhartha Saxena @siddsax]

Anthropic onboarding day: Michael Scott introducing Karpathy like he just signed Wemby in free agency.

[Siddharth @dash_third]

Gemini integration on Android Auto is so terrible. It can't do even the basic tasks properly. How are these products even getting to production 🤡

[Siddharth @dash_third]

Incentives. Here, it's getting promoted.

[Saurabh Kumar @drummatick]

I have loved Google all my life, but there’s a particular pathology I’ve been observing over the last 3 years and it’s starting to bug me.

Google keeps creating overlapping AI products with confusingly similar names.

Back in 2024, it was Gemini, Gemini Advanced, Gemini Pro/Ultra subscriptions, Gemini API, Google AI Studio, Vertex AI, etc.

Today it’s Gemini Spark, Google Antigravity, Vertex AI, Gemini API, Google AI Studio, the Gemini app, and more.

I would love to see a coherent product umbrella like OpenAI or Anthropic have built. Instead, I’m often left looking at product names that sound similar but serve very different use cases.

[Siddharth @dash_third]

@meteoraqueen1 @sidin Hahaha. My bottle was from Oct 19, 2018 at the Emirates.

[Siddharth @dash_third]

Talking to friends I haven't spoken to in years at 2 AM. All of us celebrating together, reliving the ghosts of two decades. Unreal this. Thank you, Arsenal 🏆

[Trev @Tokyo_Trev]

People are congratulating me like I was part of the squad, and I am saying thank you like I played minutes 😭

[Arsenal @Arsenal]

This belongs to all of us.

[Siddharth @dash_third]

Arsenal ♥️🏆🥲

20+ years of wait and heartbreak ends!

[Siddharth @dash_third]

🤯 That should add a few hundred billions to the valuation

[Andrej Karpathy @karpathy]

Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.

[Siddharth @dash_third]

Update after 12 matches. Accuracy isn't any better than a coin flip.

Model handles team-level structure fine. Home advantage, bowling gaps, form. 92% of outcomes land inside the band.

But it's simply not equipped to deal with one player deciding the game - Finn Allen 93(35), Starc 4/40, Marsh 90(38).

Let's see if it can adapt and learn for the remaining matches.

[Siddharth @dash_third]

Fun research experiment I'm running through this IPL.

5 IPL matches in. Polymarket seems to overprice narratives (dew, form) and underprice boring venue-specific base rates (head-to-head at the ground, chase win rates, bowling matchups). Matches 4 and 5 especially.

Based on a multi-agent system that I built to predict match outcomes. Fully autonomous on a VPS, no human in the loop and self improving via Autoresearch (thanks @karpathy sensei).

n=5 so grain of salt. Repo is public 👇

[Siddharth @dash_third]

Another AI slop giveaway phrase - X dressed up as Y. Seeing everywhere 😏

[Siddharth @dash_third]

Such confusing language. You are drastically reducing the usage limit for Claude -p effectively 😒

[ClaudeDevs @ClaudeDevs]

Starting June 15, paid Claude plans can claim a dedicated monthly credit for programmatic usage.

The credit covers usage of:
- Claude Agent SDK
- claude -p
- Claude Code GitHub Actions
- Third-party apps built on the Agent SDK

[Siddharth @dash_third]

Brutal on Diane Green!

[Internal Tech Emails @TechEmails]

Sam Altman's group chat with Satya Nadella

November 21, 2023

[Siddharth @dash_third]

This is a fantastic write-up on product management. Everything is on point 🔥

[Tom Verrilli @tdrobbo]

https://t.co/1VFZzDZdJL

[Siddharth @dash_third]

@sahilpatwa 😄 I cannot have unread tabs open, so chalk it off to that