the engineer who built Claude Code just dropped a 28-minute video on how to write prompts that actually work I've seen $300 courses that don't cover what he shows in the first 10 minutes CLAUDE.md files, memory shortcuts, parallel sessions, prompting patterns all in one video and completely free... works whether you're a developer, a beginner, or someone who's been using Claude for months based on this, I put together 18 things you can copy and use in Claude today full guide in the article belowshow more

From someone living in Claude all day: CLAUDE.md and memory shortcuts did more for my output than any prompt-engineering trick this year. Most people still treat the model like a search box. Treat it like a teammate with amnesia and the workflow changes overnight.

isn't this video from almost one year ago? -> please, don't get me wrong, content is still relevant 👍🏼 but people could confuse it with recent content on claude code from Code w/ Claude London 2026 ->

just watched it and now i'm mad at myself for prompt-engineering like a caveman for the past year. the CLAUDE.md memory thing alone is worth the 28 minutes. feels like going from threshold pace to actually knowing your VO2 max zones.

Thanks. This IS useful for real beginners. To be honest, I learned these things by trial and error and asking Claude about the best approach. Still, this saves time. Your next article might be about using skills. Both as overall approaches e g. formatting, or as mini exec code to undertake a specific, but repetitive action. You could also talk about pitfalls. The worst (for me) was to allow Claude Cowork to run on and do whatever it thought best. That included erasing folders. I tell it to stop after every action and confirm a next step. No deletions EVER! I cull files and folders manually. FWIW my process is to research using Perplexity, dump into NotebookLM to pull out substance, reduce noise, then use Claude (which is MUCH smarter) to help structure my work. If you keep scut work off paid sites, then using AI becomes viable.

Your prompt: "You are full stack coder with 20 years of expericence" My prompt: "I am full stack architect with 38 years of experience" There is a difference! ;D

So the large language model CAN'T interpret your intent and you have to PROGRAM it with specific syntax to get a desired outcome? Isn't that an admission of failure?

CLAUDE.md files alone changed how I use Claude daily. Free 28-minute video doing more than most paid courses is the kind of resource the whole community needed. 🧠

the gap between watching it and doing it is where the $300 lives

How anyone even comprehends any of this is commendable 😂😂😂

we’re entering the era where knowing how to talk to AI properly is starting to feel like a real engineering skill

What? It absolutely has cross-thread memory.

People overcomplicate prompt engineering so much Half the time the best results come from: being specific, showing examples, and patiently refining the output over and over again til it works

CLAUDE.md isn't a prompt template. It's a shared brain. The real pattern is building one canonical instruction set that multiple agents inherit, so they coordinate without you orchestrating every session. That's the architecture shift most people miss in the video.

The parallel sessions tip is underrated. I've been running 3-4 Claude Code sessions simultaneously for different modules — cuts development time significantly when you split the context properly.

This is gold. AI tools are reshaping how GTM and growth teams work in Web3 too. Having used AI heavily in my BD workflows — from prospect research to content creation — the difference between those who use it well vs. superficially is massive. Prompting is a skill. Great share!

Biggest problem i have with claude is running out of tokens in 1 or 2 prompts, even though i pay for the privilege of using claude i hit the low limits extremely fast

It's so useless I paid and started a project then one day it forgot everything and left me hanging to start the project again on my own after it failed to remember

Hold up, a free 28-min video better than $300 courses? I'm skeptical but intrigued. Saving this, especially the memory shortcuts part.

@TheOriks

Most people are still asking Claude random questions. Meanwhile others are building full workflows with it. This video explains the difference.

We're entering a weird era where knowing how to talk to AI is becoming a real skill. Not because prompts are magic, but because better context creates better output.

The very first point is wrong .. even in the same project when we start a new chat it doesn’t remember anything previously discussed which is a major drawback compared to ChatGPT , there are pros using Claude .. it is good at creating financial models , suggesting ideas

@AnatoliKopadze crazy how a free video can outdo those pricey courses. feels like a hack lol

@grok So. The set-up is complete. Whats the idea? Humans: fuck.

are we not over prompt engineering? i thought it's an ancient topic by now.

the real skill isn't the prompts though, it's knowing when to let the model fail and pivot. most people just keep tweaking the same approach til it works. that's the actual bottleneck

Thanks! Just watched the video CLAUDE.md is a game changer. Your 18 tips are gold

Prompt courses are the new incense. Everyone wants the secret words, but most of the work is still attention arranged honestly.

already tested parallel sessions - it's the kind of thing that sounds small but completely changes your workflow

We’re entering the era where knowing how to talk to AI is becoming a real engineering skill.

The funny thing about prompt engineering is that the best tips are often boring: give context, give examples, define constraints, repeat until it works.

Ah, you wrote this based on the engineer’s insights. I thought you came up with the steps based on your own experience. Still great knowledge and well put together but it makes me feel like a super genius because I have all but 3 items living and working in my Al chats and I came up with them based on 2 yrs of use in my own workflows. This tells me I am on the right track! Thank you!

Will check this

Thanks Anatoli, this is gold Just watched the video and the CLAUDE.md memory setup alone is a game-changer. Been treating Claude like a fresh intern every new chat — no wonder my projects felt scattered.Already copied a couple of those 18 tips. The parallel sessions idea is going to save me so much time.Appreciate you breaking it down with the article too. More valuable than most paid courses.