fadi

706 posts

fadi banner
fadi

fadi

@fadieacc

https://t.co/28dMJWaTxR • AI & SaaS building • regularly sharing high signal tips about product growth 🔔

Katılım Mayıs 2022
1.2K Takip Edilen4.6K Takipçiler
Sabitlenmiş Tweet
fadi
fadi@fadieacc·
if people knew 1) what Chinese OSS models will be dropped next year 2) the level of software-side optimization happening right now to squeeze massive compute into smaller and smaller machines we wouldn't price this market the same way OpenAI doesn’t hold current valuation multiples Claude doesn’t either grok is playing a different layer entirely not model vs model, but infra vs infra whoever controls fabs + power at scale sets the ceiling for everyone else fast forward 20 years, two real players: @Google @xai everyone else fighting compression, margins, and china in a race to the bottom
English
0
0
1
298
fadi
fadi@fadieacc·
it's called - AI Shrinkflation 😅 Anthropic is dynamically reducing thinking budget based on GPU load. You’re paying per-token for a model that gives you fewer tokens of reasoning when servers are busy. Every AI provider is already doing this silently. We’re tracking it with our AI Shrinkflation Index -> lmmarketcap.com/trackers/nerf-…
fadi tweet media
English
0
0
0
90
fadi
fadi@fadieacc·
it's called - AI Shrinkflation 😅 Anthropic is dynamically reducing thinking budget based on GPU load. You’re paying per-token for a model that gives you fewer tokens of reasoning when servers are busy. Every AI provider is already doing this silently. We’re tracking it with our AI Shrinkflation Index: lmmarketcap.com/trackers/nerf-…
fadi tweet media
ℏεsam@Hesamation

AMD Senior AI Director confirms Claude has been nerfed. She analyzed Claude's session logs from Janurary to March: > median thinking dropped from ~2,200 to ~600 chars > API requests went up 80x from Feb to Mar. less thinking and failed attempts meaning more retries, burning more tokens, and spending more on tokens > reads-per-edit dropped from 6.6x → 2.0x. model stops researching code before touching it. > model tried to bail out or ask "should i continue" 173 times in 17 days (0 times before March 8). > self-contradiction in reasoning ("oh wait, actually...") tripled. > conventions like CLAUDE.md get ignored because there's less thinking budget to cross-check edits > 5pm and 7pm PST are the worst hours, late night is significantly better. this means the thinking allocation is most likely GPU-load-sensitive.

English
0
0
0
130
محمد المرزوقي
محمد المرزوقي@mohd_almarzooqi·
أهالي دبي مع هالاجواء أنصحكم تزورون مشاتل الورسان . كوفي مهباش استغل الفرصة وفتح له فرع في مشتل قرين ترند
محمد المرزوقي tweet mediaمحمد المرزوقي tweet mediaمحمد المرزوقي tweet mediaمحمد المرزوقي tweet media
العربية
10
35
411
46.3K
The Spectator Index
The Spectator Index@spectatorindex·
Trump says his country is 'ready to use the full economic might of the US' to strengthen Hungary's economy.
The Spectator Index tweet media
English
803
494
5.2K
955.4K
fadi
fadi@fadieacc·
@mattjay I actually want to see more news from AI labs on security, it just showcase how much almost entirely everyone is dumb towards these stocks and we need a bigger dip to buy
English
0
0
0
196
fadi
fadi@fadieacc·
we've been cooking something for a long while 🫡 will share more soon
fadi tweet media
English
1
0
1
107
fadi
fadi@fadieacc·
I think this is still HUGELY underappreciated, but time is clearly on the side of @xai @SpaceX (Terafab), let me explain: • they owns distribution through X, at scale! (@elonmusk can just push a notification to +600M people to try a new model) • they own future compute + power generation with data centers in space, at scale! • Grok 4 already crush benchmarks, at 2M context window, when they haven't even begun yet! if you control both distribution and energy at scale, you're playing a different game entirely. everyone else can keep competing on models and benchmarks in the short term, but that's not where this ends.. long term, this consolidates super hard. as I always say in the US, the end-game winners in AI are Google and xAI. mark my words
fadi tweet media
English
0
0
2
840
fadi
fadi@fadieacc·
@xai @SpaceX @elonmusk I mean Grok 4 is already converging hard, and they haven't even started yet..
fadi tweet media
English
0
0
0
62
fadi
fadi@fadieacc·
I think this is still HUGELY underappreciated, but time is clearly on the side of @xai x @SpaceX, let me explain: • they owns distribution through X, at scale! (@elonmusk can just push a notification to +600M people to try a new model) • they own future compute + power generation with data centers in space, at scale! if you control both distribution and energy at scale, you're playing a different game entirely. everyone else can keep competing on models and benchmarks in the short term, but that's not where this ends.. long term, this consolidates hard. as I always say in the US, the end-game winners in AI are Google and xAI. mark my words
English
1
0
0
104
fadi
fadi@fadieacc·
Gemma 4 31B is still SO underrated!
fadi tweet media
English
0
0
0
63
fadi
fadi@fadieacc·
future AI dominance litmus test: these are consistently the top countries visiting our AI leaderboard lmmarketcap.com China is very focused on AI, their population is all in early signals.
fadi tweet media
English
0
0
0
82
fadi
fadi@fadieacc·
@elonmusk what would it take to lead cybersecurity and anti-abuse at @xai ? I’m builder and CISO running cyber for a platform with millions of users, I’ve done it at scale - I can share a full proposal with your team on how to turn this around along with my actual background
English
0
0
0
158
fadi retweetledi
fadi
fadi@fadieacc·
"Overloaded" 🤠 Dario needs more compute, @dylan522p was right - Anthropic will hit a wall on compute sooner or later, OpenAI is far scaled up and even have back up And Grok will pull off the biggest lever in the century, they're cooking it slowly
fadi tweet media
fadi@fadieacc

Claude is down again

English
0
1
0
404
fadi
fadi@fadieacc·
@TheAhmadOsman quite it is. are you referring to it in the most optimized settings or just as is
English
0
0
0
263
fadi
fadi@fadieacc·
for context, I already bought both domains anyways
English
0
0
0
73
fadi
fadi@fadieacc·
traffic has been growing and I'm pushing this app toward what it’s supposed to become. based on this tweet from @balajis I'm mostly going to migrate away from aimodelsmap.com to either 1. lmmarketcap.com 2. llmmarketcap.com which is best?
Balaji@balajis

We need AI market cap. Meaning: Coinmarketcap, but for AI usage. What’s the best coding model today? The best image model? Best video model? It just constantly changes. So: rank on the basis of Google trends, usage by your friends, and other variables.

English
1
0
0
440
fadi
fadi@fadieacc·
@yourealazyfvck pretty interested but the "Pretty competitive so come correct" part kinda messed up the fun for me.. injured historically so I prioritize competitive yet chill games, not losing a leg again lol
English
2
0
3
4.8K
David
David@yourealazyfvck·
If you’re in Dubai and you’re free on Wednesday nights we run a 5/6 a side football (soccer game) every week at 8pm Pretty competitive so come correct Hit me for details
English
130
18
756
181.3K
fadi
fadi@fadieacc·
@damengchen wow! what do you use for cold outreach, same?
English
0
0
0
340
Damon Chen
Damon Chen@damengchen·
Used mailcheep.com to send a broadcast campaign to 16k contacts, and it only cost $1.66 from my AWS bill. ✌️
English
53
19
507
66.2K

Keşfet

@mohd_almarzooqi @BasedMediaMan @spectatorindex @mattjay @xai @SpaceX @elonmusk @dylan522p