rahmanyoo

Running off rails never worked out. I personalized differently for different LLMs. To truly get LLMs to suit your personal style takes time and often have to redo with every new major version release. Gist of common instructions are to make it ask clarifying questions, TLDR of every response at the top, To be direct and Rewards for showing true confidence scores for every response (this becomes critical in long conversations).

Glad to see real outcomes from small teams.

> Once Claude familiar with your repo (ask it to make copious notes), async coding is the next level of productivity. I now ask it to write a lot of the repetitive code paths.

Can you share what is your approach to context engineering this ?

Also what AI tools you use on daily basis ?

My confidence in Mozilla got renewed after watching their CTO's interview on early days, early decisions, solving bottlenecks, Google being a Frenemy, major bug in BigZilla etc.,

Suffice to say it's learned experience makes one a Senior?

Excellent reminder in age of AI slop. In context of article, it's soul-less.

Isn't it interesting, In a supermarket we are hesitant to buy products from unfamiliar brands but in App/Play store, anything goes.

great! i’ll check it out.

Nokia-ish moment.

Roomba brought commercial application of IoT & robotics into our homes, much earlier than others. Failed to keep up with innovation.

Just as Nokia pioneered early smart phones and didn’t keep up with times.

Apple eerily feels like going down similar path with lack of AI adoption and botched Siri.

Congrats Ran! Nice and clean landing page as well.

It would help if landing page shows some examples, makes purchase decision easier.

If you’re looking for a reviewer, i’m happy to!

As a founder, i find great joy in learning about Day 1 story on anything.

Most fascinating Day 1 story i’ve come across is Cosmic Ladder talk by math genius Terence Tao.

Thanks for sharing.

This maybe an unpopular stance, when i read to learn something new or a fictional story, i noticed AI summaries misses the nuances of good writing that our brain intuitively picks up.

For some reason (maybe there is a psychological explanation), information retention is better when i read from the source rather than an AI summary.

Reward of learning something new in its original form, vastly outweighs ‘time saved’ by AI summaries.

What works for me, pass 1 is scanning through original text, pass 2 is digging into areas that needs deeper understanding and pass 3 is assessing understanding using AI summaries.

i have not. Just a personal preference to write specs myself because i want as much control at early planning stages.

The Beginning of Infinity by David Deutsch is in same category. I have not finished reading it yet, at best i go 5 pages in one sitting. Jam packed with fascinating facts.

Not sure if there is an industry standard. Langchain is a decent place to start if starting from scratch. If you’re building to learn, i wouldn’t worry so much about standards, pick something that has decent community & docs.

Good luck!

I tried both ways. Letting design emerge incrementally adds vibe-coded technical debt.

At minimum prepare project specs with features as .md file. I found claude to be useful for this task.

If you’re using Cursor, pick a template from here. https://cursor.directory/rules