pasmichal

pasmichal

@pasmichal.com

Biggest tech eng. AI practitioner, context is king. 5-9 builder with kid, wife & hobbies. Dragging aviation out of the tech stone age http://notamify.com 🦋 whichstarterpacksamiin.com - wondering where those follows come from?

594 followers169 following316 posts

I'm considering making a career change to AI/ML engineer I've been doing web development for 15+ years, there's not that much new things to learn for me. Not only I'm not stimulated anymore, I think I'm stuck. NGL Changing careers sounds scary at 40yo

I played myself to use standing desk more. My chair broke down and I deliberately didn't replace it. I tried sitting on a stool instead, but it's not comfortable for any periods longer than 15 mins for me. Now I'm looking forward to work standing up.

"Success is when your children want to be with you when they're adults" I'm not a fan of sharing those deep thoughts, but this one hits right in the heart. Raising kids is not always fun or pleasant, sometimes it stinks (literally), but when it's good... it's the best feeling in the world

Just bought a new domain... 😬 aislopstop.com AI/vibe coding is a thing and it's going to grow. We need some tooling and education how to develop secure, reliable apps with AI.

AI slop stop #1: Use AI to write tests first, then implementation Writing tests is easier than ever, there is no reason for projects not to be extensively tested nowadays. AI will produce nonsense sometimes and having regression tests is your first line of defense.

You wouldn’t believe how many very skilled engineers at FAANGs are dismissive of AI “It makes things up,” “It has no idea,” “It makes mistakes.” Today, I heard from one senior engineer that he “deliberately avoids anything AI".

Vibe coding will be some day a reliable of writing software, espacially in established frameworks with good tooling (Laravel comes to mind). It is not there yet tho and pretending it is leads to results like this 👇

Self reposting is dumb and it makes you look like an attention whore Before I hit ‘follow,’ I check profiles. If I see self-reposts everywhere, it’s an instant turn-off

This quote from Russell hits even harder in social media era “The whole problem with the world is that fools and fanatics are always so certain of themselves, and wiser people so full of doubts.”

Journaling is great for mental health, personal growth, and well-being. I tried it multiple times. Always stopped within the first 2 weeks. Here's what worked for me this time: - use a very basic tool, I use jrnl.sh - make AI summarize your week in review and send it to you

"How I use AI" is a golden content idea 👌 Useful for both a creator and consumer, not like shitposting. YT might be a better platform for such content tho

The "good engineer" is killing my progress on the side project I always try to get things done properly because I know how severe consequences can be... Gotta tap into that reckless me and ship more

One big benefit of scheduling posts is that it stops me from refreshing engagement metrics to check if I'm going viral (spoiler: never went viral) delayed gratification > instant gratification

If you are not using Gemini Flash 2.0 you're missing out - It's cheap as dirt - It's very capable - It's a easy to use as any other LLM Try aistudio.google.com for prompt engineering

So far, my experience with Grok 3 has been a bit underwhelming Used Grok (with deep search) as a tutor for my ML learnign path and it led me to a dead end and 3h lost. GPT 4o did much better in explaining and providing step by step code samples / instructions.

"Don’t Build AI Products The Way Everyone Else Is Doing It" video aged very well https://youtu.be/bRFLE9qi3t8?si=q30arAcznYrHuxJD 1. If you can do smth without AI do it 2. Out-of-the-box LLMs is not the end 3. Train your use-case specific models

Notamify tech stack: Quart (python) Next.js Firebase PostgreSQL Google cloud GPT You probably never heard of quart, huh? Me neither before I joined 😅 but it's good!

Even after 15+ years of engineering I still have to remind myself that it's OK to fuck up as long as you can fix it. Don't let the fear of messing up prevent you from shipping 💪 We are all imposters 😅

The number of apps someone shipped / launched should not be a metric of success. No matter the time frame. How much success does the app have is a metric. By success I mean 💰, it's business after all 🤷‍♂️

First impressions after giving Zen browser a try: 1. Seamles swtich, all the data import works great 2. UI less cluttered - more space for the website itself 3. ♥️ Workspaces is a banger - I can have different set of pinned tabs whether I'm working on notamify.com or just browsing

Prioritization is haaard 😫 There are parts of the system I’d love to improve for dev experience, but they don’t directly move the business needle. Balancing what excites you vs. what pays the bills? The eternal founder struggle.