Abdelrahman Soliman

Perceptr Repos: github.com/perceptrio/per… - the python backend github.com/perceptrio/per… - the UI Dashboard github.com/bebofit/percep… - the web sdk for recording

I’m pivoting from Perceptr. Here is the technical deep dive, hopefully you learn from it. For the past year and a half, I’ve been working on a product that analyzes session replays to help teams understand what users are doing before they complain. The idea was simple: Session replays already contain a goldmine of data: DOM changes, clicks, scrolls, console logs, network requests, frustration signals, and user behavior. Instead of manually watching recordings, could we use AI to detect bugs, usability issues, and patterns automatically? Technically, this was one of the most interesting things I’ve built. I learned how session replay tools actually work under the hood, studied open-source projects like RRWeb and PostHog, built a web SDK, streamed events in batches, used S3 pre-signed URLs, compressed sessions with gzip, experimented with video conversion, and eventually replaced the expensive video-based approach with a much cheaper pipeline using algorithms, Playwright screenshots, and human-readable event summaries. The cost went from around $0.30 per analyzed session to almost $0.01. That was a big engineering win. But the bigger lesson was not technical. The product worked, but selling it was much harder than I expected. For many teams, this was more of a “nice to have” than an urgent painkiller. And that’s probably the most important startup lesson I’m taking from this: Great engineering does not automatically mean a great business. You can solve hard technical problems, optimize costs, build a clever architecture, and still be wrong about the market. I’m still proud of what I built, and I’m open-sourcing it after some cleanup so others can learn from it or maybe even use it inside their own teams. Perceptr didn’t become the company I hoped it would be (this is just a pivot, but I love to be dramatic haha). But it gave me a much deeper understanding of product analytics, AI workflows, infrastructure trade-offs, and most importantly, what kind of problems are worth building around next.

This week was a humbling one. I’ll be pivoting from Perceptr. Sometimes the best decision is to be honest, close the chapter properly, and move forward. I still plan to open-source Perceptr, document what I built, and maybe share some technical deep dives from the process. The technical challenge was genuinely fun, and I hope someone can learn from it one day. Now I’m back in the messy phase: figuring out who I want to serve and what problem is worth solving. A few directions I’m exploring: - Non-technical founders who need a better way to understand and manage the technical side of building a product. - Early-stage CTOs dealing with the hectic part of SOC2 and enterprise sales - AI engineers thinking about how AI gets deployed, secured, and made reliable in production. The common theme I’m noticing is that I enjoy bringing order to chaos. Whether it’s product data, technical decisions, or confusing workflows, I like helping people turn messy information into clear action. This phase can feel slow, but I’m realizing it’s not wasted time. It’s the part of building that people rarely talk about — the abyss. Now time for discovery calls I guess

@tomboutin_ lol this could be a startup idea

11pm, in bed, vibecoding a duck on a plane as a meeting reminder. We’re so back 🦆

@daniel_tian1 @photon_hq PHO

I will fly you out to sf all expenses paid you’ll get: > 3 days in sf > fully covered flights, food and housing > a chance to build with @photon_hq :) Photon Residency is looking for TOP growth, engineers and designers that reimagine agent-human interaction Comment “PHO” for the link to join us in SF (yes this is chapter 2 and we are doing it again)

@vaaselene What tips would you suggest?

New landing upgrade ✨ appreciate all the tips on Claude design and adding taste.

@vaaselene 100% I started investing in public speaking and storytelling courses. Every investor pitch, discovery call, customer sale, and convincing an employee is a story that needs to be engaging

what's the most underrated skill for a founder? i think storytelling skill 👀

@jeremybernier So sorry for this but you connect the dots looking backwards! Wishing all success

I just got laid off from Meta. Obviously it sucks to lose the income. But between the never-ending layoffs, stack ranking, etc., I'm good. Pretty convinced that when I look back at this moment a few years into the future, I'll be grateful it happened. There's a lot I want to say here, but I'll save that for another day. Either way it's been a privilege to have been a part of the ride. On to the next adventure.

@nikunj It is very painful. I have been trying to get my company up for a year and a half solo and it is humbling. Doing everything I can and question myself why not just find a job and get on with it, but the fulfillment of solving a problem and making an impact is what keeps me going

Being a founder is so freaking hard man.. chewing glass every day. If you fail, no one cares. If you succeed, you are given way more problems. Sometimes I really wonder how many founders wake up in the morning and ask themselves "is it really worth it". Starting a company is so glorified - but building, scaling and maintaining one is a different story. This is why it's so important to do this for the right reasons - solve problems you are deeply passionate about and one you'll do even if the world is against you. At least there will be light at the end of the tunnel that's constantly drawing you in on those days when you are questioning your motivation. I have never had the courage to personally start a company for this reason. Just wanted to say to the ones going through this - I see you, I salute you and I admire you 🫡

@ashleyhindle Really Codex is better than Sonnet 4.6? I dont use Opus that much mainly for planning or something is it good for UI as well (without a skill)?

Hope codex limits reset shortly so I don't have to get another subscription🤞💸 Claude mostly dead to me atm, except for Opus 4.7 being infinitely better at UI 🧑‍🎨

@julesrosenberg @FarzaTV @grok How should I not scroll X, it is how I connect with people now.

5 things @farzatv does differently > never checks email (his agent does it) > takes analog notes at the gym, turns them into apps > uses screenshots and songs as inspiration > built a personal wikipedia of his own life > never scrolls X, uses @grok instead

I would say 1- Caching prompts dramatically saves a lot especially system prompts, tools definition getting sent each time 2- smartly routing which context should be sent 3- not using MCPs blindly 4- use strong models for planning and cheaper models for executing (cheaper models for low-stakes tasks)

Realizing prompt engineering isn't that important for controlling AI costs Pipeline architecture is: - cheaper models for low-stakes tasks - smaller inputs - smarter sync cadence the prompt is the last thing to optimize

@ashleyhindle Overfitting 101 lol, an ai model cant be left in charge i guess

Messed up my planning stage and was asked 17 questions in a row before I gave up and told it to just work 😂 Who knew this stuff wasn't as easy as it looks?! Crazy

@sunniekapar Ghostty is 55.7GB and Cursor 41GB? damn proof of shipping features

we launched people to the moon with kilobytes of ram, but i cant even run my nextjs server without my computer blowing up

@foliofed I was just watching a youtube video about this from Mobbin extremly useful youtu.be/Qsq-Sj_rojU?si…

I used to think a good SaaS onboarding is about collecting data It's really all about creating useful artifacts, ensuring a good first-time user experience, and delivering value Most of the profile questions you'd ask can be figured out from the domain anyways

@imittalishika Yeah, but it takes a toll on you. I am learning to appreciate resting and sleeping as much as work. Still very hard to do lol

When you wake up in the middle of the night because sleep feels like a waste of time, and it feels like it’s keeping you away from your work. You are lucky, because you truly love what you do. I am lucky :))

@hthieblot yeah still there but for some weird reason I am still going haha

Everyone loves the idea of being a founder. Until you’re 1 year in, broke, have zero traction, and your family is asking when you’ll get a “real” job.

@adamc0dez @ycombinator Congrats!! wishing all success

Excited to share I just got into @ycombinator! It took me a while to find my login info to my email linked to my account but once I did the forgot password flow worked flawlessly.

@hthieblot Perceptr watches your users and proactively understands their behavior so product managers can take precise decisions perceptr.io

Can this new X algo find incredible builders/founders? If that’s you tell me what you are building.

This week, I started with 2 more discovery calls for Perceptr. The signal is getting clearer: this workflow might be more of a vitamin than a painkiller. I also had a mentorship call with @arvidkahl , author of The Embedded Entrepreneur, and he introduced me to a different way of finding niche ideas: I listed groups of people I like working with or learning about: product managers, engineering managers, CTOs, founders, coffee enthusiasts, book readers, ecommerce store owners, video gamers, AI engineers, and more. Then I started scoring each group across 4 dimensions: 1. Affinity Do I genuinely enjoy these people? Do I want them to succeed? Would I be happy serving them for the next 10 years? 2. Opportunity Where are they “screaming”? What are they complaining about on Reddit, in communities, in reviews, and in forums? What tools already exist? Where are they still underserved? (Chatgpt deep research is amazing here) 3. Willingness to pay Do they already spend money solving problems in this space? Are there products, subscriptions, services, or events proving that this market pays? 4. Market size How many of them are there? And if I start with a very narrow niche, can it expand into something bigger over time? Each niche gets a score out of 40. Choose top 3 -> First one is your idea and you have two more when you need to pivot One important Note: Use AI for research, not judgment. The judgment still has to come from you. it has to be subjective Still working on this sheet happy to share it and all resources 💪 youtu.be/Pde9iazFWq0?si…