Sharekh (@Sharekh_) header
Sharekh (@Sharekh_) profile picture

2X Founder @CleverXHQ and now building an energy company. In love with 🏍️ and 📚

San Francisco, US
Joined March 2017
Photo by Sharekh (@Sharekh_)Photo by Sharekh (@Sharekh_)Photo by Sharekh (@Sharekh_)Photo by Sharekh (@Sharekh_)Photo by Sharekh (@Sharekh_)Photo by Sharekh (@Sharekh_)Photo by Sharekh (@Sharekh_)Photo by Sharekh (@Sharekh_)Photo by Sharekh (@Sharekh_)
BYD builds a better car than Tesla and still can't sell in America. Kimi will learn the same lesson. The frontier isn't won on benchmarks, It's won on who is allowed to compete. OpenAI and Anthropic already know this.
2
226
It’s funny to watch Founder: here is my VC horror story What does the VC do 🤷 VC: Here’s my GP horror story.
5
151
Opus is driving me crazy right now, ruining my late night mojo! Nerfed say what you will
2
152
I’m loving what @AnthropicAI is doing to fix sycophancy in their models.
3
78
🤯
This quoted post is unavailable.
49
Founders pour everything into building. No guarantee it pays off. The one thing you can control is to build along with people who share your values. Otherwise, it’s not worth it.
1
48
What an incredible time to build things. Don’t get distracted by grifters. Focus on what matters, customers, team, partners and most importantly what you are to become.
2
42
All of my thoughts have moved to the terminal. I’ve almost stopped using a pen and paper, not sure if this is a good thing.
2
30
The gap between "I wish I could build this" and "I built this" collapsed. Every founder with a backlog of ideas just got years of bandwidth back. The ones who ship will win. The ones who wait will watch.
5
49
"We tried ChatGPT for our legal briefs, but..." This is how 90% of my customer calls start. The "but" is always the same: hallucinations, inconsistent outputs, responses that sound like they've never seen a law firm. You don't have an AI problem. You have a customization problem. Generic AI is like hiring a brilliant intern who's never worked in your industry. Smart, but doesn't know your language, standards, or workflows. Fine-tuning changes that. We took one legal client from "technically accurate but professionally useless" to "indistinguishable from senior associates" in 4 weeks. Stop accepting 70% accuracy as "just how AI works." With proper fine-tuning, 95%+ reliability isn't just possible—it should be standard.
3
7
151
Soham Parekh was the guy Sam Altman was talking about the one-person billion-dollar company. And we just stopped that from happening. This makes Peter Thiel right, we are not innovating fast enough.
2
134
Reading ‘hard thing about hard things’ again. And the one thing I keep thinking is that it’s not good enough to just know where the biggest problems in the business are. But having the courage to proactively go after them and choosing your pain, pain is inevitable either way.
3
85
Intellignece is overrated and courage is underrated
4
74
It will be super interesting to experiment if the operator can be trained to use business apps
RL keeps cooking 🤣🤡🫡 “Deep research was trained using end-to-end reinforcement learning on hard browsing and reasoning tasks across a range of domains.”
2
148