Angehefteter Tweet
Sup*
1.7K posts
Sup*
@aibenchmarking
AI | Model Benchmarking & evals | Hackathons & personal projects | Building curative intelligence | I run an agency btw but this profile ain't about that
Vibing Beigetreten Ağustos 2024
351 Folgt78 Follower
@richiemcilroy Why do I forget all this before buying more credits!!!!!!
English
claude is always down
just stick it on a $5 hetzner server and it's good to go forever 😤
English
Imagine a vibe coder creating the rotation logic and suddenly it twists your whole neck
Cooper Mitchell@homegymcoop
When the late, great Arthur Jones sold Nautilus (original inventor of bodybuilding machines) he created MedX. MedX made some of the most ridiculous, yet “effective” equipment of all time. This is one of those machines, the Cervical Rotation Machine.
English
@KaranVaidya6 Dude
1000 businesses spending over $1 million p.a. is both diabolical and crazy cool at the same time ⚡
English
We are one of the 500.
If you want to be part of the elite, DM me.
himanshu@himanshustwts
One for the history books.
English
The energy in the agency just spiked in the last hour and this was the reason.
And now I see designers going gaga for @sketch
Sketch@sketch
🆕 Things you’ve been asking us for 🆕 Selection colors, independent borders, corner smoothing controls, a new eyedropper with Color Variable support, and 150+ improvements and fixes. Here’s what’s in our latest update — Dublin 🧵 ↓
English
If this sounds like you, I'm hiring.
Shane Levine@theShaneLevine
People ask me how has AI changed the design process? Just look at @tryramp Product Designer job description
English
Claude Cowork is powerful but right now nothing compared to OpenClaw
It’s a race. May the best product win.
Hosea@hoseakidane_
@garrytan Gary, have you tried claude cowork for the same tasks you would want openclaw for? A lot of the features overlap now like using iMessage, scheduling, using computer use etc
English
@aibenchmarking Pick anything related to OS models. Consulting, fine tuning, synthetic dataset preparation, on premise deployment.
Anything.
English
I'm doing that already.
> took OS model
> fine tuned it on (80M dataset)
> now running 24/7 on my macbook
> with 98% accurate tool calls
> it design its own workflows
> can talk, research, automate, save
> and much more
Launching soon.
Marc Andreessen 🇺🇸@pmarca
Magical OpenClaw experiences that use frontier models cost $300-1,000/day today, heading to $10,000/day and more. The future shape of the entire technology industry will be how to drive that to $20/month.
English
@aibenchmarking Tool calling, instruction following, and low tier reasoning.
English
I've been fine-tuning open source models lately, (again)
and I'm loving it.
It's the first time I'm feeling that I'm in control, models are learing super fast and on benchmarks they are beating Sonnet 4.5, Gemini 3.0 easily.
SLMs (Small Language Models) are the real unlock.
They can run on any consumer device. No API, No Cloud, No provacy issues.
And they can beat SOTA models on niche specific use cases. I've done that myself.
Enterprise will through bundles of cash on specialized SLMs because of the control and power it brings.
The only bottleneck is cloud compute, (GPUs are not available on major inference providers).
I'm considering buying my own GPU rig because GPU prices will definitely go up.
Also I felt that there's too much technical stuff that no one is simplifying for beginners (how to fine-tune models exactly)
So, I'll be doing that. I'll share my findings, struggles and few wins here.
Time to come out of stealth. Join the ride.
English
@dakshpixelup "AI is the new hype"
People have been saying the same thing for years now. What in your opinion is the reason for the industry to flourish even on a cash burn rate 👀
English
300B poured into startups last quarter
80% of it went to AI
thats more AI companies competing for the same enterprise contracts than any quarter in history
with everyone selling the same promise, here’s the truth nobody wants to admit:
The best product doesn’t always win.
The best perceived product does.
Enterprise doesn't pick features.
They pick confidence.
And confidence = brand.
Product opens the door.
Brand closes the deal
English
@skirano @MagicPathAI Ah, when creating a component we do have the option to select Mobile devices to adjust the canvas size, but on preview (new tab) it opens in desktop mode. I am suggesting to get a dedicated Mobile prototype installed in which the user can try out the component.
English
@aibenchmarking @MagicPathAI You can send any design to your phone and you'll see it, or do you mean something else?
English
We pushed a cool update in @MagicPathAI.
Now when sharing a link to a public file, on mobile you see a "feed" of all the prototypes in there.
It's such a cool way to share projects with your team and clients, and truly only possible on our platform.
English