Hemant Rai

450 posts

Hemant Rai banner
Hemant Rai

Hemant Rai

@Hemantisme

Training AI with just enough compute and a lot of patience.

Pune, India Katılım Eylül 2009
434 Takip Edilen132 Takipçiler
Sabitlenmiş Tweet
Hemant Rai
Hemant Rai@Hemantisme·
🧵 Training an AI Model on a Low-End GPU (RTX 3060) Think you need an A100 to train AI? Think again. I trained a model on 22GB of text using an RTX 3060 (12GB VRAM) with smart optimizations. Here’s how I made AI training work on a budget. 👇 #AI #ThriftTrainer #LowCostAI
English
1
3
5
659
Hemant Rai retweetledi
Chessfeed AI
Chessfeed AI@ChessfeedAI·
The queen was hanging. Black took it. But the queen was bait. Bxf7+ forces the king to e7, and Nd5# ends the game. Before grabbing material, check your king.
English
0
1
2
93
Hemant Rai retweetledi
Chessfeed AI
Chessfeed AI@ChessfeedAI·
The queen had confidence. The position wanted more pieces. Chess Blunder Stories #1: attacks need pieces, not just confidence.
English
0
2
2
33
Hemant Rai retweetledi
Chessfeed AI
Chessfeed AI@ChessfeedAI·
You reviewed the mistake. But did you practise it? chessfeed Study can turn moments from your own games into practice positions. Go back. Find the better move. Build the pattern. chessfeed.ai
English
0
2
3
54
Hemant Rai retweetledi
Chessfeed AI
Chessfeed AI@ChessfeedAI·
Stop guessing what to look at. Study on chessfeed helps surface the signals in a position: king safety, loose pieces, pawn breaks, threats, weaknesses, and plans. Notice first. Then decide. chessfeed.ai
Chessfeed AI tweet mediaChessfeed AI tweet mediaChessfeed AI tweet mediaChessfeed AI tweet media
English
0
2
2
70
Hemant Rai retweetledi
Chessfeed AI
Chessfeed AI@ChessfeedAI·
Your games now create your practice. Blunders, mistakes, and missed mates from your own games can become positions you solve again. Replay the mistake. Find the better move. Build the pattern. chessfeed.ai
Chessfeed AI tweet media
English
0
2
2
28
Hemant Rai retweetledi
Chessfeed AI
Chessfeed AI@ChessfeedAI·
Ask the coach about the exact position you are studying. Plans. threats. candidate moves. mistakes to avoid. Grounded in the board, not a generic chess tip. chessfeed.ai
Chessfeed AI tweet media
English
0
2
3
30
Hemant Rai retweetledi
Chessfeed AI
Chessfeed AI@ChessfeedAI·
One position can lead to completely different games. That is why Study lets you branch candidate moves and compare what changes. Try one idea. Go back. Try another. The goal is not just to find a move. The goal is to understand the position. chessfeed.ai
Chessfeed AI tweet media
English
0
2
3
42
Hemant Rai retweetledi
Chessfeed AI
Chessfeed AI@ChessfeedAI·
The hard part in chess study is not always finding a move. Sometimes it is knowing what to think about. Study Feed gives your study direction: what to notice, compare, practise, or ask the coach next. chessfeed.ai
Chessfeed AI tweet mediaChessfeed AI tweet mediaChessfeed AI tweet mediaChessfeed AI tweet media
English
0
2
3
46
Hemant Rai retweetledi
Chessfeed AI
Chessfeed AI@ChessfeedAI·
A new way to learn chess is now live on chessfeed.ai. Study turns your games and positions into guided learning sessions with signals, insights, branching, practice, and coach help. Built for understanding, not memorisation.
Chessfeed AI tweet media
English
0
2
2
85
Hemant Rai retweetledi
Chessfeed AI
Chessfeed AI@ChessfeedAI·
Your games can now teach you back. The new Study experience is live on chessfeed.ai. Turn your games into guided learning sessions with signals, insights, branching, practice, and coach help. Built for understanding, not memorisation. chessfeed.ai
Chessfeed AI tweet media
English
0
4
5
101
Hemant Rai retweetledi
Chessfeed AI
Chessfeed AI@ChessfeedAI·
This is exactly how we think about it. Not GM-level, plenty of room to improve, but aiming to explain ideas in a way that’s genuinely useful to a large part of the chess-playing population. Reducing fluff is very much part of the roadmap.
Srinath Narayanan@srinathchess

@anishgiri @DavidHowellGM I wouldn't describe it as 'low quality'. Rather, it still contains a lot of fluff. Obviously it is not at GM level and there is a lot of room for improvement, but I think it already explains in a way that is relevant for significant numbers of chess population.

English
0
1
1
115
Hemant Rai retweetledi
Chessfeed AI
Chessfeed AI@ChessfeedAI·
@anishgiri @srinathchess @DavidHowellGM That’s exactly the concern we share. In this case, when pushed with an incorrect premise, the coach rejects it and justifies why it doesn’t hold. That kind of pushback is what we’re aiming for, though there’s still a long way to go.
Chessfeed AI tweet media
English
0
2
2
115
Hemant Rai retweetledi
Anish Giri
Anish Giri@anishgiri·
@ChessfeedAI @srinathchess @DavidHowellGM This looks like typical ChatGPT behavior to me. Now try to tell him some wrong ideas. Will he also be as receptive to them? e.g. Say that rook frees up the square for the Kf1-Ke1 maneuvre, bringing the king closer to the defense of the c3 pawn (to make it absurd).
English
1
1
1
234
Hemant Rai retweetledi
Chessfeed AI
Chessfeed AI@ChessfeedAI·
@anishgiri @srinathchess @DavidHowellGM Fair point 🙂 We tried your critique directly in Chessfeed and pushed the coach on the concrete ideas. This kind of back-and-forth, challenge, correction and refinement is exactly what we’re building Chessfeed for.
Chessfeed AI tweet media
English
1
1
3
246
Hemant Rai retweetledi
Srinath Narayanan
Srinath Narayanan@srinathchess·
@anishgiri @DavidHowellGM I wouldn't describe it as 'low quality'. Rather, it still contains a lot of fluff. Obviously it is not at GM level and there is a lot of room for improvement, but I think it already explains in a way that is relevant for significant numbers of chess population.
English
1
1
4
824
Hemant Rai retweetledi
U Mumba
U Mumba@umumba·
Time to show your love for the Mumba Masters! 💚 Tell us who your favourite player is & stand a chance to win exciting prizes! ♟️ Powered by @ChessfeedAI #upGradMumbaMasters #TechMGCL
U Mumba tweet media
English
0
3
5
189