Joined January 2019
- Tweets 270
- Following 71
- Followers 8
- Likes 132
Photos and videos
If you could permanently recruit ONE person as your AI teammate:
Elon Musk
Steve Jobs
Jensen Huang
Sam Altman
Who are you choosing?
1
23
Apr 19
How did Higgsfield hit $200M ARR in just 9 months?
Here’s what actually worked:
1. MVP: solve ONE real bottleneck
They didn’t try to build everything.
They identified a key pain point in AI video creation: motion control.
1
31
Apr 19
3. Monetization: aggressive packaging
They pushed high-commitment plans:
→ 2-year plan
→ “unlimited” plans for new models
Not cheap.
But aligned with power users.
References:
– Check early versions on Web Archive
– Traffic growth on SimilarWeb aligns with product updates
1
14
Apr 19
Takeaway:
They didn’t win by better models.
They won by better packaging distribution.
8
18 Nov 2025
Many people believe that being the first automatically means winning.
But the data tells a different story:
Only half of the first-mover advantage is real — the other half can be a trap. ⚠️
I wrote a deep-dive explaining:
why 47% of pioneers fail 💣
substack.com/profile/3995759…
36
1 Dec 2024
Bayesian learning is about examples, especially unique ones. Access to diverse data makes theory meaningful. 📊🔍 #Bayesian #DataScience #Learning
35
29 Nov 2024
To ensure outsourced products meet expectations:
1️⃣ Tech selects partners.
2️⃣ Product evaluates based on past work.
Clear roles = better results! 🚀👩💻 #Outsourcing #ProductDev #Teamwork
27
29 Nov 2024
Bayesian thinking in investment: Focus on failures, build a "failure database," update beliefs, and refine decisions to minimize risks. 🚀 #Investing #Bayesian #Entrepreneurship
28
28 Nov 2024
【💡 Using AI to Boost Communication】
I use Plaudnote to record and analyze convos, improving clarity. AI summaries highlight strengths and areas for improvement, speeding up skill development. #AI #Communication
20
13 Oct 2023
30 Mar 2023
GPT-4 should know that my statements about DL never rising above "curve fitting" is not an opinion, nor a "common concern" but a mathematical theorem. You cannot answer causal questions from passive data alone, unless you enrich the data with: (1) causal assumptions or
1/3
1
90
28 Sep 2023
How to Build a Brand on TikTok
Organizationally:
1. The top executive must get involved to set an exemplary role.
2. 20% of the people have a premium, the rest are at regular prices.
1
65
28 Sep 2023
Improving Content Efficiency, Making it a Mathematical Problem:
1. Label videos and analyze data. Elements like storyline, ambiance, background music, close-ups, and camera movements should be considered.
2. Establish a workflow, optimizing each part of the process.
56
14 Sep 2023
Here's the translation of the provided text into English:
"How to utilize Bayesian principles to refine daily behavioral norms.
Objective 1: In the first 5 minutes of each Pomodoro timer, systematically and automatically choose the task to be performed.
1
50
14 Sep 2023
2. Prioritize based on importance (ratings: 1, 2, 3), urgency (in relation to the next deadline), and relevance (whether it can be done in 3 Pomodoro cycles, its relevance to the previous Pomodoro task, and if the relevant resources are ready. Ratings: 4, 5, 6).
1
36
14 Sep 2023
3. Sum the ratings in reverse and sort to determine the task to be done.
4. Use 4 Pomodoro cycles as a unit.
Testing period: 2 workdays 2 off-days.
Cons: Lacks alternative hypotheses, making comparisons challenging.
Pros: It's detailed and is a first step."
41