_mad_eye_

for transparency, here's what we're learned about what will make contents go viral so far (human effort and quality control):

1. YC: trending topics; sharing failure in public 2. Career decisions → Document your career pivots; they resonate because everyone's navigating transitions. 3. Personal story, e.g. moving country → because they signal ambition and momentum—and you can tell them again in 6 months. 4. Good hooks in general, e.g. cultural differences → Cultural differences are natural hooks; they're unexpected and inherently shareable. 5. Real event sharing → Real events (conferences, launches, wins) get amplified because people cheer for actual momentum, not vibes. 6. Showing hustling: meeting new people, having lots of coffee chats etc. 7. Sharing personal success (Awards ≥ events = personal win) 8. an image of yourself will 3x impression 9. a nice personal story will 3x impression 10. what goes viral on one platform can go viral on another platform 11. what goes viral will go viral again.

We're building Posting Machine, we make sure your linkedin posts go viral, grow your account, and find your ICPs from those that engaged with your contents.

We're looking for early customers and design partners to help us iterate fast. If you are a b2b saas founder who values linkedin as a sales pipeline, we'd love to talk to you!

https://cal.com/posting-machine/30min

continuing -- so you don't have to spend time on Linkedin and do chores like ideating, posting, replying to comments, analyzing ppl's profiles etc.

This is the tldr of social selling: If someone's liked your LinkedIn content, then if you reach out to them it's no longer cold outreach. It's ~warm outreach, response rate will go up by 80%.

We don't post AI slops because we spend a large amount of time learning what's going to work and what doesn't. And we put a lot of human care into the posts for our clients to make sure that it works.

While most content agency asks for $3000-$5000 per month, we price at $499/mo, but we're offering an early bird of $149 per month.

Me and Hank both used to work at YC startups, and we also have YC company founders as our initial clients. Would love to hear what you think!

yeah vibe engineered

doesn't work with papers without HTML content yet!

Have LLMs sip through them for you according to your taste!

If you're interested in this solution, please join the beta testing! :)) It takes 5 usd/mo, I offer 30 day money feedback guarantee.

Initial feedback from my researcher friend is that the results are very accurate (relevant).

Let me know what you guys think!

demo video: https://www.youtube.com/watch?v=kcYQ3oHpIFc

A little bit, I'm doing some public test for the time being.

So the other day I was thinking about whether a reasoning LLM can learn our interest better than those machine learning personalization services, see [1]. And then I decided to built one.

You can try it here https://arxiv.ianhsiao.xyz , feedbacks are welcomed!

[1] https://www.reddit.com/r/MachineLearning/comments/1iol1qd/d_...

I'd recommend Gilbert Strang's Introduction to Linear Algebra and 3B1B's calculus and linear algebra series. They focus on the intuition part of math, which makes it fun and interesting. After that do some problem sets, that should give you a solid foundation to understand ML papers with the help of things like ChatGPT.

It adds explanations after technical or specialized terms

I thought of the idea while doing psychology-related research, since I'm very unfamiliar with its English terminologies.

Useful for 1. News reading: Helps readers quickly understand complex topics in news articles, such as economics, science, or politics, without needing to search for explanations. 2. Cross-disciplinary reading/research: Enables readers and researchers to grasp unfamiliar terms from fields outside their expertise, fostering interdisciplinary learning and collaboration. 3. Language comprehension: Provides real-time clarifications for jargon-heavy content, helping non-native speakers or those with limited exposure to specialized terms better understand the material.

Example of effect:

Before: The quantum entanglement phenomenon demonstrates nonlocality, challenging our understanding of quantum mechanics and EPR paradox.

After: The quantum entanglement (a phenomenon where particles remain connected regardless of distance) phenomenon demonstrates nonlocality (the ability of particles to instantly influence each other at any distance), challenging our understanding of quantum mechanics and EPR paradox (a thought experiment questioning quantum mechanics' completeness).

I'm still on the fence about whether this is something people actually want. Do you mind if I learn a bit more about your current workflow with case studies?

The tool helps you compare different voices by the topics on the same topic(s) efficiently.

I use it to study how different companies approach problems like hiring, product strategy, org design etc. But as of now I'm not sure what are the user segments I should directly talk to.

Backstory of the idea first (why), followed by how to use it.

Backstory: I am a reader of Lenny's Newsletter, and I was reading the "How {company} builds product" series. But reading alone is not enough, the ideas are easy to read and easy to forget.

Learning these things for me, means comparing how different companies make decisions under the same topic, like hiring, building culture, org structure, etc. I want to compare their answer to the solution, and their reasoning if possible. So I have to deconstruct each article into different single theme chunk, and then sort it by themes (hiring, product strategy,..), then I ask questions about each theme ("what worked for ..., why?") to get useful results ("YC mentality + senior talent"), and then finally, I get to "organize" the information to better build my understanding on the topic.

Real learning comes from 'reading' and 'organizing', the rests (deconstructing, sorting, finding themes) are steps that prepare the information that's helping you to 'organize'. And I think the value of this tool (Reconstruct) is to help you skip the manual preparing part (it takes a shit ton of time and energy).

How to use it: (if it breaks, just hit refresh) 1. Upload some documents of the same topic, ideally newsletter, high quality articles, podcast transcripts etc. Personally I also think this works well with YC startup library too. This only accepts .txt and .md now 2. Click 'deconstruct' to break it down into multiple chunks. It should take less than 10 seconds. 3. Click 'identify themes' to learn which topic that chunk is talking about. It's good when you can find repeated themes (colors) in the articles. This should take around 30 seconds, depending on how many chunks there are. 4. Click 'sort' to view the chunks by their theme 5. Select some cards you are interested in, and click 'Extract by'. Enter what you want to extract from the cards, and you will get the results next to it. 6. To understand the results better, you can go through step 5 multiple times, until you get all the useful information you want. You can then switch mode to connect mode (cmd +2) and connect the cards together to form a learning.

If you think this can help you, also feel free to contact me at [email protected]

You can try the paper "Attention is all you need" too if you want to :)) https://proceedings.neurips.cc/paper_files/paper/2017/file/3...