Jack (@4nsdp) profile picture

Databases

Joined January 2024
Pinned Tweet
Software engineers: Do you know why JWTs start with ey? πŸ€”
142
37
35
1,496
674,646
Jack retweeted
Been learning about LLMs for a while now. Spent some time this week building a tiny GPT from scratch, training it on 448K software arch Q&A examples. 256 token context, 8K vocab, ~2.9M params Not very accurate yet, but a fun learning colab.research.google.com/dr…
1
3
122
Jack retweeted
sam altman is kinda right
🀝 Paid partnership
3
2
5
294
sam altman is kinda right
🀝 Paid partnership
3
2
5
294
Jack retweeted
With Ai, Agents & Robots being spammed right left & center. The new moat is human. Build cafe, resturants, malls or even tools by human for human! πŸ“ˆ
1
2
90
Web devs: Do you know why URLs have two slashes (//) after http:? πŸ€” Not 1. Not 3. Why 2?
77
20
1
507
242,398
A new term is echoing through AI circles these days: Loop Engineering 🧢 Do you know what it actually means? ⬇️ For the last 2 years, the workflow looked like this: You write a prompt πŸ‘¨β€πŸ’», AI gives an answer πŸ€– You write another prompt πŸ‘¨β€πŸ’», AI gives another answer πŸ€– You were in the loop πŸ‘¦ Now the idea is different. You build a system that: - Finds work πŸ“‹ - Prompts the agent πŸ€– - Checks the results πŸ§ͺ - Updates memory πŸ“ - Decides the next step πŸ”„ Without waiting for you. You're no longer prompting the AI. You're designing the loop that prompts the AI. The leverage point has moved. 🎯 Prompt Engineering taught us how to talk to models. πŸ—£οΈ Loop Engineering is about teaching models how to keep working when we're not talking to them. 😢 And that's a much bigger shift than most people realize. Have you tried it? 🧐
23
1
30
10,732
Coders: Do you know why every program, in any language, starts with main()?
81
11
4
125
22,049
Is writing SQL queries off the top of your head still a flex?🫠
41
4
2
51
12,773
Ever wonder why QR codes have 3 squares in the corners? Why 3, not 4 or 2? πŸ€”
41
2
1
52
45,166
Ever wondered how LLMs understand negation if they read text one token at a time? 🧢 Consider this sentence: ⬇️ I don't think the movie was terrible. As the model reads: πŸ‘‰ I don't think... it starts expecting a negative opinion? Then it sees: πŸ‘‰ terrible Now things look even more negative. But the actual meaning isn't strongly negative at all. In many conversations, "I don't think it was terrible" is closer to: βœ… It was okay. βœ… It wasn't that bad. So how does the model get this right? πŸ₯Ά The trick is that LLMs don't lock in meaning as each token arrives. πŸ’Ώ Every new token can change how previous tokens are interpreted. 🧢 When the model sees words like don't, not, never, or hardly, it uses attention to connect them with other relevant words in the sentence. 🀯 In this case, attention helps the model understand that don't modifies the idea of terrible, changing the overall sentiment. πŸ‘¨πŸΌβ€πŸ³ Meaning doesn't come from individual words. It comes from how words attend to each other. πŸ’‘ And that's why a tiny token can completely flip the meaning of a sentence. 🧠 Bcz Attention is All You Need. 🀯
27
1
37
88,794
Interviewer: You say "Hey Siri πŸ‘‹πŸΌ" and it responds. Is your iPhone always listening? How does it work?
34
2
51
46,406
What’s a bigger career dead-end: refusing to use AI πŸ¦–, or letting an agent build a system you can't debug? 🧱
56
5
57
4,479
is AI the biggest human invention since we learned to harness electricityπŸ’‘?
34
39
3,070
What’s a bigger red flag in an interview: a dev who can't prompt a model 😢, or a dev who can't explain the code? 🚩
42
2
54
4,313
We've all heard that modern LLMs have billions of parameters. 🧢 Some of the latest models are rumored to have over a trillion. 🀯 Now think about that for a second. If a model has 1 trillion parameters, and each parameter takes 4 bytes, that's 4 TB just to store the weights. πŸ₯Ά And every time you ask the model a question❓, those weights need to be available to the GPUs performing inference. πŸ“ So how is it possible for AI companies to serve millions of requests without needing racks of GPUs per conversation? ⬇️ The answer is quantization (mapping floating point weights into lower-bit integers like 8-bit or 4-bit using scale + zero-point to approximate values during inference) Just like instead of full value for pi we just limit to 3.14. πŸ’‘ Most weights don't actually need 32 bits of precision. You can often represent them using 8 bits, 4 bits, or even fewer, while preserving most of the model's quality. πŸ€– Suddenly that 4 TB model becomes 1 TB. Or 500 GB. Or less. Same model. Same knowledge. Far less memory and bandwidth. The funny thing is that many people think AI inference is limited by compute. πŸ’Ώ In reality, a huge part of the challenge is simply moving model weights around fast enough. 🚒 Quantization works because the model usually doesn't need perfect numbers. It just needs numbers that are accurate enough. And that simple idea is one of the reasons trillion-parameter models are practical at all. πŸ«±πŸ»β€πŸ«²πŸΌ
36
3
54
107,304
be honest, Is "vibecoding" just a fancy word for not knowing how your own app works anymore? 🫠
26
2
36
1,754
Spent the weekend shipping with the latest MiniMax M3 model πŸ€– for $0 A few observations: ⬇️ - Very good at spec-driven development - Strong backend engineering capabilities - Great at troubleshooting and debugging - Frontend is okay so far. Have you tried it? What's been your experience? πŸ‘€
47
2
60
3,138
Unpopular opinion: A boring micro SaaS solving a painful, expensive problem is 100x better than an AI wrapper that's just a fancy prompt. 🧢 What happened to just building genuinely useful software?πŸ€·β€β™‚οΈ
29
1
40
2,251
Do you know how a reasoning model actually works? πŸ€” Does it have some special reasoning engine hidden inside? 🧩 Not really. ⬇️ Imagine asking two AI models: "What is 47 Γ— 23?" The first one instantly replies: 1081 The second one says: 47 Γ— 20 = 940 47 Γ— 3 = 141 940 + 141 = 1081 Same answer. βœ… But this time you can see how it got there. That's essentially what a reasoning model is. It's not a new architecture. It's the same LLM, trained to think step by step before answering. How? 🧠 Chain of Thought β†’ learn to break problems into smaller steps. 🎯 Reward Models β†’ reward good reasoning, not just correct answers. πŸ”„ Reinforcement Learning β†’ reinforce reasoning paths that lead to better outcomes βœ… Self-Critique β†’ review and fix its own mistakes before responding The result? Instead of making a fast guess, the model learns to spend extra tokens thinking. Same brain. Different training.
26
1
37
2,034