Light

6.9K posts

Light banner
Light

Light

@lightfidreams

Technologist | Entrepreneur | Building @gameable_world | Relentlessly optimistic

Katılım Kasım 2021
2.4K Takip Edilen2.2K Takipçiler
Morgan
Morgan@morganlinton·
@paraschopra Love seeing this, and also totally wild you’re doing this on an M3 w/36gb of ram. Also, how did you get 36gb? I thought it was only 32 or 48 that you could get?
English
2
0
4
724
Paras Chopra
Paras Chopra@paraschopra·
I tried replicating fully agentic coding on my laptop and I'm impressed! My setup - Qwen 27bn (4bit quant) huggingface.co/unsloth/Qwen3.… - Pi coding agent pi.dev - CCO sandboxing for yolo mode github.com/nikvdp/cco My Mac M3 has 36GB ram. It's surreal to see a local model read prompt, follow it flawlessly and then build a fully self-contained html/css page with zero errors. I got about ~20 tokens/second, and on this prompt quality similar to Haiku! But... this is running locally :)
English
59
58
1.1K
84.1K
Tibo
Tibo@thsottiaux·
It’s the little things that matter, what are some small papercuts you have noticed in Codex? We’ll fix as many as possible in the next week.
English
1.9K
57
2.3K
240.9K
jeremy
jeremy@jercarin·
@badlogicgames I mean this with no malice but genuinely who uses kilo code?
English
3
0
2
2.2K
Light
Light@lightfidreams·
@mattpocockuk Besides harnesses are general purpose this is TS only what you want so…
English
0
0
0
7
Matt Pocock
Matt Pocock@mattpocockuk·
One thing I wish harnesses did by default: When opening a file, FIRST pre-compile the file and extract only the type signatures and comments for that file (with tsgo this would be instant). Then, if you want to see the implementation, only unwrap the functions you're interested in. Essentially .d.ts for the first step, .ts for the second. Would save a ton of tokens and allow agents to explore more aggressively.
English
46
11
410
34.4K
Light
Light@lightfidreams·
@RhysSullivan If you defer your thinking you also lose your value completely as an engineer.. biggest (personality) skill issue ever
English
0
0
0
22
Rhys
Rhys@RhysSullivan·
not having to type is really nice, but i think i want to go back to manually writing the code myself and more leveraging LLMs for research and understanding of the codebase it's just too easy to defer you thinking today and end up in a bad state
Rhys@RhysSullivan

from my experience, even the best models (Opus 4.6, 5.4 xhigh / 5.3 codex) cannot write good code today without an amount of work that is equivalent to just doing the work myself am excited for a world where they can, but in the current state i have very low trust in them

English
53
39
895
55.4K
Rhys
Rhys@RhysSullivan·
from my experience, even the best models (Opus 4.6, 5.4 xhigh / 5.3 codex) cannot write good code today without an amount of work that is equivalent to just doing the work myself am excited for a world where they can, but in the current state i have very low trust in them
English
278
88
2.8K
321.3K
Light
Light@lightfidreams·
@levelsio 36 mins is NOT a summary brother
English
0
0
1
122
Light
Light@lightfidreams·
@mitchellh @adamhjk If you are in the loop and carefully iterate and specify you can really get the primitve done with agentic engineering, but you have to know what you want to be seeing from the output, not depend on the model to tell you the shape
English
1
0
0
170
Mitchell Hashimoto
Mitchell Hashimoto@mitchellh·
It also lets agents build stable software. The primitives being proven, solid and well written pieces of tech that agents honestly struggle at on their own is what enables the higher levels to be more sloppy (and FINE). With a really careful harness and lots of prompt engineering and tool dev maybe it’s possible to build something solid like libghostty but I think it’s not with today’s models and the token cost would be astronomical. I think that’s true of many building blocks. Some have called it “cached tokens” but I’m more pessimistic about it I actually think current models are incapable of it. But the higher layers don’t need as much attention to detail.
English
3
3
58
17.9K
Adam Jacob
Adam Jacob@adamhjk·
The most useful and interesting property of software in the Agent era so far is its ability to adapt to your circumstances. @mitchellh talked about this in his post about the "building block economy", framed through the lens of what grows fastest.
English
1
0
18
2.7K
Light
Light@lightfidreams·
@gabriberton Wdym dead code a simple lint rule is enough
English
0
0
1
113
Gabriele Berton
Gabriele Berton@gabriberton·
Vibe coding creates lots of dead code. Run this often. You're welcome --- Delete all dead code. Use ruff and vulture ---
English
158
265
6.4K
690.8K
Light retweetledi
@melissa
@melissa@melissa·
i think a lot about how to get a high agency kid here’s what’s working +
English
11
58
1.5K
215.2K
limitlesstack
limitlesstack@limitlesstack·
the vaginal lining is actually a great delivery mechanism for a variety of drugs goes directly into the bloodstream. no first pass metabolism. max potency. royal jelly applied intra-vaginally outperformed estrogen for female libido. the queen bee has the highest sex drive of any creature in the hive. she mates once. with everyone. she's fed exclusively on royal jelly. her drive. her mood. her initiation. all of it goes up. give her royal jelly. you'll know when it's working. dutch origin, ships worldwide. by @aestheticprimal this is where i get mine: aesthetic-alimentum.com/products/royal… not medical advice.
limitlesstack tweet medialimitlesstack tweet media
_aestheticprimal_@aestheticprimal

Go down on her with Royal jelly on your tongue it improves her quality of life, libido and microbiome

English
25
90
1.7K
358.2K
Light
Light@lightfidreams·
@benln Ok ya sure
English
0
0
1
93
Michael
Michael@michael_chomsky·
Here’s an idea.md for anyone who isn’t scared to build an ambitious product: Someone’s going to make 100-1000M dollars building a self-updating personal knowledge base that syncs with imessage, twitter, email, chatgpt/claude/claude code/codex messages. This knowledge base will have an MCP to be accessible from anywhere. You’ll be able to edit it like notion, style it to your liking, and set rules about how data is organized. Once it gets mature enough, it can even proactively suggest things that will improve your life, as it knows everything about you. Unlike some memory systems, this will just be files so completely observable. The closest thing to this that exists is Mintlify’s KB and Notion, but both are more for enterprises than individuals. Just paste this into Claude Code, ask it to perform a socratic interview about ambiguities (or just use gstack), clank for 2 days, get Chamath and Karpathy as your first two customers, and do YC. You’ll have competition at some point but it won’t matter because you’ll be better at UGC and paid ads.
Chamath Palihapitiya@chamath

This may be a dumb question but I’ll ask it here anyways: I can’t find a good way for my various AI chats to automatically sync its conversation history into a structured knowledge base. So that as I update various chats from time to time and refine context, my knowledge base automatically grows with this new info.

English
74
28
610
103.5K
Light
Light@lightfidreams·
@alightinastorm heh nice but i wouldn't call this a tool damn
English
0
0
0
79
robot
robot@alightinastorm·
bro casually called it a study and dropped a threejs mega banger lool
English
4
9
378
33.7K
Light
Light@lightfidreams·
@daradoescode @tibor_tee In older design mode u can’t click on an alement and prompt a change inline, you’d have to do it in separate agent chat
English
0
0
1
13
Dara A.
Dara A.@daradoescode·
@tibor_tee I thought design mode was already in cursor tho
English
2
0
7
336
Dara A.
Dara A.@daradoescode·
Can someone give me a run down of what cursor 3 actually is? Is it just cursor glass but now out of beta?
English
10
0
32
5.6K

Keşfet

@morganlinton @paraschopra @thsottiaux @jxnlco @jercarin @badlogicgames @mattpocockuk @RhysSullivan