Ibrahim Ahmed
@atbeme
Followers
5K
Following
26K
Media
785
Statuses
8K
building @inference_net | forever @fdotinc
San Francisco, CA
Joined May 2014
If you're spending over $25,000 on OpenAI/Gemini/Claude, you need to talk to our team at @inference_net We just finished training a custom model for a leading consumer app that cuts their latency by 70% and their costs by over 50%
4
4
67
who's on the OpenAI Codex-Slack feature, this is so funny -- we can do better ya'll come on now
1
0
5
Spot on
I have a long-running list of these surprises for first-time Founders and CEOs. Here are a few that werenβt on Brianβs list. 1. Your calendar becomes everyone elseβs calendar. Every gap gets filled. 2. The hardest conversations are never on the calendar. They start with βGot a
0
0
2
Cerebras 4x boost, easy
The era of fast coding models has begun: Cognition SWE-1.5: ~950 tok/s Cursor Composer-1: ~250 tok/s Haiku 4.5: ~140 tok/s
0
0
3
Absolute genius, I was talking about exactly this idea about two years ago when I did my last move
SellRaze (@SellRaze) uses AI to sell your stuff from just a pic. It identifies it, prices it, and lists it across marketplaces automatically. What used to take 30+ minutes now takes seconds. https://t.co/WQVEySzxV9 Congrats on the launch, @JeffMao_ & @TylerMa_!
1
0
7
if you're a software developer and you're not using AI, you're ngmi you'll eventually give up and adopt the tech, or you're going to get absolutely cooked by folks who are just 10x faster than you you still need strong CS and software eng fundamentals tho, slop is still slop
1
0
18
codex >>> claude code for super hard problems claude code >>> codex for stuff that's more visual
3
0
12
Using custom-trained LLMs and > 1k 4090s to visualize 100k scientific research papers in latent space π DM me for early access π
229
354
4K
Can someone from @hyperbolic_labs help us out with some instances that we're running on the cloud? Looks like Discord is pretty inactive and the support bubble chat is really rough in getting issues resolved
3
0
5
Today we're releasing Schematron: 3B and 8B models that extract typed JSON from HTML with frontier-level accuracy. We built Schematron to be the ultimate πΈπ°π³π¬π©π°π³π΄π¦ for data extraction: β 50-100x cheaper than GPT-5 β ~10x lower latency β Both models outperform Gemini 2.5
9
15
128
We worked with the @calai_app team to train a custom calorie estimation model for their core user flows. Today, this custom model powers 100% of their traffic, serving millions of users worldwide. The model is: - Better than GPT5 - 3x faster - 50% cheaper read it and weep π
50
21
716