-

510 posts

-

-

@angilly

Katılım Eylül 2008
2.4K Takip Edilen2.4K Takipçiler
Sabitlenmiş Tweet
-
-@angilly·
@angilly" target="_blank" rel="nofollow noopener">macaw.social/@angilly
ZXX
0
0
5
456
-
-@angilly·
@mattshumer_ Analogous to what diffusion models do!
English
0
0
0
109
Matt Shumer
Matt Shumer@mattshumer_·
Forget vibe coding. It's time for Chaos Coding: -> Prompt Claude 3.7 Sonnet with your vague idea. -> Say "keep going" repeatedly. -> Watch an incredible product appear from utter chaos. -> Pretend you're still in control. Lean into Sonnet's insanity — the results are wild.
English
305
568
11.4K
1.2M
-
-@angilly·
@vikasreddy Are you using any tools to help with this or hack something together homegrown?
English
1
0
1
51
Vikas Reddy
Vikas Reddy@vikasreddy·
Using LLM + full codebase context with Rust is so awesome! The compiler catches any dumb small stuff and you can use the LLM to iteratively fix compile errors. Makes it high probability things "just work" when you go to actually run the code!
English
2
0
5
581
-
-@angilly·
@adamcohenhillel Cool once those are in place it’d be cool to see how far down the quantization rabbit hole you can go and still maintain accuracy.
English
0
0
0
168
ADAM
ADAM@adamcohenhillel·
@angilly Not official ones yet, will do that in a few days - but did run my own eval for my use cases (first aid medical responder)
English
1
0
0
33
ADAM
ADAM@adamcohenhillel·
So you can actually train a small model to be an expert in a specific domain by leveraging a larger, more capable model to teach it! I made a simple Colab notebook (open sourced) to fine-tune llama-3-8b on a specific knowledge generated by the huge Llama-3-405B model 👇
ADAM tweet media
English
13
83
783
85.8K
-
-@angilly·
@adamcohenhillel Did you run evals to quantify the improvement?
English
1
0
0
172
ADAM
ADAM@adamcohenhillel·
Small models are useful and important because you can run them on your own personal hardware at home (even raspi 5!) - making it much more secure, private, and cheaper. Small models for the win! Also, if you or your org need to train&run specialized models - my DMs are open
English
1
2
15
1.6K
Jaana Dogan ヤナ ドガン
If I were organizing office hours to brainstorm about LLM developer tooling, would you be interested? 👇
English
86
8
364
41K
Jacqueline Ros Amable
Jacqueline Ros Amable@jackie_ros·
I just found out my home insurance did not renew due to climate change. I knew this was coming but I didn’t think this year 😭
English
1
0
0
212
Star Simpson
Star Simpson@starsandrobots·
ChatGPT UI all broken for everyone or just me?
Star Simpson tweet media
English
2
0
1
749
-
-@angilly·
@sama @ChatGPTapp being able to search my chatgpt history is becoming a non-trivial productivity unlock. Can we get that worked into the roadmap? Or can you expose the history API so we can build our own search w/o needing to trust 3rd party browser extensions?
English
1
0
1
82
- retweetledi
Mike Johnston
Mike Johnston@MikeJohnstonCO·
I know many Denverites will observe 4/20 today. Weed appreciate if you celebrate responsibly in the Mile High, and don’t drive if you’ve been using. We need a joint effort to keep everyone safe!
Mike Johnston tweet media
English
45
9
77
19K
James Kelley
James Kelley@MuchRockness·
How to use Vision Pro: Put it on Wait 30 seconds Realize it's off Hold a button for 5 seconds Hold other button 5 seconds Hold other button 7 seconds Hold other button 7 seconds Wait another 10 seconds Light gray glow appears How to use Quest 3: put it on
English
34
11
181
29.9K
-
-@angilly·
@nielr1 @petergyang What’s AI? Sounds like I should try it out. Can you fax me some instructions?
English
1
0
0
17
nielr1
nielr1@nielr1·
@petergyang I think its going to take a lot longer for the general population to know how to interact with AI to get much out of it. We've been programmed to search with the expected answer being a link or a document, not the answer itself. That'll take time to relearn.
English
1
0
1
53
Peter Yang
Peter Yang@petergyang·
Feels like we're in a sweet spot right now where AI is already incredibly useful but still patiently answers all the dumb questions and prompts that we ask it
English
4
1
23
8.6K
- retweetledi
Sam Altman
Sam Altman@sama·
fk it why not 8
English
1.5K
798
16.9K
6.2M
- retweetledi
MC HAMMER e/acc
MC HAMMER e/acc@MCHammer·
Salute and congratulations to the 710 plus @OpenAI team members who gave an unparalleled demonstration of loyalty, love and commitment to @sama and @gdb in these perilous times it was a thing of beauty to witness. Salute 🥂🥂🥂🥂🥂🥂🥂
English
88
204
2.3K
1.1M
-
-@angilly·
@VytasBu @thestephse Screenshot shows msg saying “I was reading your website” Haha no you weren’t.
English
1
0
1
11
Vytas Bu
Vytas Bu@VytasBu·
@angilly @thestephse How is personalisation in cold outreach a lie? :D I've been using it for ages.
English
1
0
0
87
Stephan Seider
Stephan Seider@thestephse·
I made a personalised LinkedIn Outreach automation that runs on auto-pilot. Helped me save 100+ hours already. It creates hyper-personal messages based on their company websites. Interested? Like this post & comment "prospect" and I'll DM you (Must be following me).
Stephan Seider tweet media
English
3
0
5
272
-
-@angilly·
@yoda Oh no what
English
0
0
0
43
-
-@angilly·
@nielr1 Mirror the dark/light mode setting of the OS. Some apps do dark mode without this and it drives me crazy.
English
1
0
1
42
nielr1
nielr1@nielr1·
Twitter brain... What features make products feel like they are designed for engineers/devs? ex: Dark Mode, extensive keyboard shortcuts, what else..
English
1
0
0
279
Johannes Schickling
Johannes Schickling@schickling·
Is there an easy way for me to have `my-project.localhost:80` instead of `localhost:3000`?
Johannes Schickling tweet media
English
402
125
1.7K
638.4K