A M Y T • Think AI Outside Box 🇮🇳 (@BoyInDaBox89) header
A M Y T • Think AI Outside Box 🇮🇳 (@BoyInDaBox89) profile picture

A M Y T • Think AI Outside Box 🇮🇳

@BoyInDaBox89

Building Sales AI Personalization Agent at Scale || Autodidact || Polymath || Talks about Quant, LLMs,AI Agents, Backend & Crypto | All opinions are personal

Joined October 2009
  • Tweets7.2K
  • Following1K
  • Followers262
349Media
Photo by A M Y T • Think AI Outside Box 🇮🇳 (@BoyInDaBox89)Photo by A M Y T • Think AI Outside Box 🇮🇳 (@BoyInDaBox89)Photo by A M Y T • Think AI Outside Box 🇮🇳 (@BoyInDaBox89)Photo by A M Y T • Think AI Outside Box 🇮🇳 (@BoyInDaBox89)Photo by A M Y T • Think AI Outside Box 🇮🇳 (@BoyInDaBox89)Photo by A M Y T • Think AI Outside Box 🇮🇳 (@BoyInDaBox89)Photo by A M Y T • Think AI Outside Box 🇮🇳 (@BoyInDaBox89)Photo by A M Y T • Think AI Outside Box 🇮🇳 (@BoyInDaBox89)Photo by A M Y T • Think AI Outside Box 🇮🇳 (@BoyInDaBox89)Photo by A M Y T • Think AI Outside Box 🇮🇳 (@BoyInDaBox89)
Imagine practicing at 38 years old then...🤣
Owen Carey (@owenthcarey) profile pictureOwen Carey
@owenthcareyJul 2
Practicing LeetCode at 32 years old.
114
A M Y T • Think AI Outside Box 🇮🇳 retweeted
Samay (@Samaytwt) profile picture
Samay
@SamaytwtApr 6
Unpopular opinion: If you can't code, you can't vibe code.
Image from Samay (@Samaytwt)
396
251
53
4,201
143,940
A M Y T • Think AI Outside Box 🇮🇳 retweeted
Mert Yuksekgonul (@mertyuksekgonul) profile picture
Mert Yuksekgonul
@mertyuksekgonulJan 22
How to get AI to make discoveries on open scientific problems? Most methods just improve the prompt with more attempts. But the AI itself doesn't improve. With test-time training, AI can continue to learn on the problem it’s trying to solve: test-time-training.github.io…
Image from Mert Yuksekgonul (@mertyuksekgonul)
25
167
54
746
389,182
A M Y T • Think AI Outside Box 🇮🇳 retweeted
Scott Adams (@ScottAdamsSays) profile picture
Scott Adams
@ScottAdamsSaysJan 13
A Final Message From Scott Adams
Image from Scott Adams (@ScottAdamsSays)
Image from Scott Adams (@ScottAdamsSays)
12,891
30,931
6,955
187,145
43,368,050
A M Y T • Think AI Outside Box 🇮🇳 retweeted
Sebastian Raschka (@rasbt) profile picture
Sebastian Raschka
@rasbt22 Dec 2025
The LLM training eras: 202x Pre-training (foundation) 2022 RLHF + PPO 2023 LoRA SFT 2024 Mid-Training 2025 RLVR + GRPO
50
143
16
1,508
167,247
A M Y T • Think AI Outside Box 🇮🇳 retweeted
Naval (@naval) profile picture
Naval
@naval27 Aug 2025
Founders can delegate everything except recruiting, fundraising, strategy, and product vision.
465
2,003
160
19,155
1,536,730
A M Y T • Think AI Outside Box 🇮🇳 retweeted
ₕₐₘₚₜₒₙ (@hamptonism) profile picture
ₕₐₘₚₜₒₙ
@hamptonism8 Aug 2025
people are nothing,
Image from ₕₐₘₚₜₒₙ (@hamptonism)
23
2,075
44
13,988
550,789
I second that!!! 😀
sankalp (@dejavucoder) profile picturesankalp
@dejavucoder5 Aug 2025
waiting for .@karpathy sensei's "let's build gpt oss in 4 hours" video
47
A M Y T • Think AI Outside Box 🇮🇳 retweeted
Naval (@naval) profile picture
Naval
@naval3 Aug 2025
In an age of stochastic compressors, be incompressible.
274
351
61
4,905
431,107
A M Y T • Think AI Outside Box 🇮🇳 retweeted
Julien Chaumond (@julien_c) profile picture
Julien Chaumond
@julien_c25 Jul 2025
BREAKING: we've partnered with @metaai and @paperswithcode to build a successor to Papers with Code (which was sunsetted yesterday) PWC, founded by @rbstojnic and @rosstaylor90 has been an invaluable resource for AI scientists and engineers over the years (and an inspiration for us to build @huggingface) We are happy to follow in their path and provide a new section of HF for the community to follow trending Papers, linked to their code implementations on GitHub 🔥 link in the next tweet ⤵️ And hat/tip @joespeez @ThomasScialom for helping with this 🥰
Image from Julien Chaumond (@julien_c)
50
217
51
1,201
180,280
startup playbook 101 😎
Garry Tan (@garrytan) profile pictureGarry Tan
@garrytan23 Jul 2025
Real taste in products is not in the eye of other famous or noteworthy builders Taste is in being able to understand customers or users, model them and make choices that delight them and solve problems for them, and be right over and over again with incomplete information
44
Sir Ramanujan is the GOD 🙏
Fermat's Library (@fermatslibrary) profile pictureFermat's Library
@fermatslibrary23 Jul 2025
Ramanujan said that he received his formulas from God
Image from Fermat's Library (@fermatslibrary)
207
A M Y T • Think AI Outside Box 🇮🇳 retweeted
Avinash Singh (@AvinashSingh_20) profile picture
Avinash Singh
@AvinashSingh_2020 Jul 2025
✅ Looking for job opportunities in India’s top funded startups? I’ve curated a list of startups in India that have raised funding in 2025, where you can apply for jobs or inquire about opportunities through cold emailing. How to Use This List? 1- Visit the LinkedIn Jobs section and apply directly. 2- On the company’s LinkedIn page, find employee data to request referrals. 3- Visit the career portal and apply directly. 4- Send them a cold email to inquire about job opportunities. You can find email addresses in the footer or on the "Contact Us" page. 🔗 Want access to this? 1️⃣ Like ❤️ & Follow 2️⃣ Repost 🔁 to help others 3️⃣ Reply “India” ✍️ 👉 Make sure to follow so I can send you the link! Once this post hits 100 likes & replies, I’ll drop the link in the thread for everyone. Don’t miss it! 🔁 Share this with friends who are job hunting in the startup space! #startups #funded
Image from Avinash Singh (@AvinashSingh_20)
137
100
1
661
48,553
A M Y T • Think AI Outside Box 🇮🇳 retweeted
DAIR.AI (@dair_ai) profile picture
DAIR.AI
@dair_ai20 Jul 2025
Top AI Papers of The Week (July 14 - 20): - Agentic-R1 - Context Rot - Scaling up RL - A Survey of AIOps - Chain-of-Thought Monitorability - One Token to Fool LLM-as-a-Judge - A Survey of Context Engineering for LLMs Read on for more:
10
144
4
982
146,613
A M Y T • Think AI Outside Box 🇮🇳 retweeted
Dev Valladares (@dev_valladares) profile picture
Dev Valladares
@dev_valladares18 Jul 2025
Infinite Wiki ⁕ Every word is a hyperlink. Every description is generated in real-time (in ~1 second) ⁕ Runs on Gemini 2.5 Flash Lite. ASCII diagrams using 2.5 Flash
106
208
67
3,027
259,030
Honest take from a mathematician on OpenAI's IMO Gold wins though Google DeepMind isn't far behind #IMO2025 #OpenAI #deepmind #LLMs
Ernest Ryu (@ErnestRyu) profile pictureErnest Ryu@ErnestRyu19 Jul 2025
Two cents on AI getting International Math Olympiad (IMO) Gold, from a mathematician. Background: Last year, Google DeepMind (GDM) got Silver in IMO 2024. This year, OpenAI solved problems P1-P5 for IMO 2025 (but not P6), and this performance corresponds to Gold. (1/10)
162
A M Y T • Think AI Outside Box 🇮🇳 retweeted
Ernest Ryu (@ErnestRyu) profile picture
Ernest Ryu@ErnestRyu19 Jul 2025
Two cents on AI getting International Math Olympiad (IMO) Gold, from a mathematician. Background: Last year, Google DeepMind (GDM) got Silver in IMO 2024. This year, OpenAI solved problems P1-P5 for IMO 2025 (but not P6), and this performance corresponds to Gold. (1/10)
52
289
41
3,822
720,241
A M Y T • Think AI Outside Box 🇮🇳 retweeted
Daniel San (@dani_avila7) profile picture
Daniel San
@dani_avila718 Jul 2025
Run this command in your terminal: claude mcp add "Deep Graph MCP" npx -- -y mcp-code-graph@latest microsoft/markitdown That’s all it takes to connect the codebase graph of any GitHub repo with Claude Code 🙌
This quoted post is unavailable.
Try viewing anyway · View on X
4
22
1
138
11,804
A M Y T • Think AI Outside Box 🇮🇳 retweeted
AVB (@neural_avb) profile picture
AVB
@neural_avb18 Jul 2025
A curriculum-ordered list of concepts to learn Reinforcement Learning: FOUNDATION - MDP - Bellman Eqn - Bellman Expectation Eqn - Value iteration - Policy Iteration - Tabular Q learning, SARSA DQNs - DQN - Double DQN - Dueling DQN - Replay buffers, Prioritized Experience Replay - On policy vs Off policy POLICY GRADIENTS - REINFORCE - Advantages - A2C, A3C - DDPG, TRPO - PPO, GAE - SAC COOL ALGORITHMS/CONCEPTS - GAIL, IRL, Imitation learning - Hindsight Experience Replay - Curiousity - Random Network Distillation - HRL - Sim to real MODEL BASED - Dyna-Q - MPC - World Models - Dreamer MULTI AGENT - POMDPs - Independent variants (I-PPO) - Decentralized vs Centralized - Centralized critic decentralized execution (MA-DDPG) - Self play - parameter sharing - value decomposition (VDIN, QMIX) - Communication rabbit hole
10
32
294
16,058
A M Y T • Think AI Outside Box 🇮🇳 retweeted
Sebastian Raschka (@rasbt) profile picture
Sebastian Raschka
@rasbt19 Jul 2025
From GPT to MoE: I reviewed & compared the main LLMs of 2025 in terms of their architectural design from DeepSeek-V3 to Kimi 2. Multi-head Latent Attention, sliding window attention, new Post- & Pre-Norm placements, NoPE, shared-expert MoEs, and more... magazine.sebastianraschka.co…
The Big LLM Architecture Comparison

The Big LLM Architecture Comparison

From DeepSeek-V3 to Kimi K2: A Look At Modern LLM Architecture Design

magazine.sebastianraschka.com
48
472
32
2,263
208,184
Load more