Activity
Mon
Wed
Fri
Sun
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
What is this?
Less
More
Memberships
Clief Notes
23.5k members • Free
Tech Snack University
18.6k members • Free
11 contributions to Clief Notes
5d •
How far do you personalize?
I've been using Claude Code and other such tools for a while now and one of the most impactful customizations I've made is telling them how I think and how to communicate with me. For example, I put simple instructions in ~/.claude/CLAUDE.md that tell Claude that I want it to always lead with the conclusion then explain, that I'm prone to spending too long perfecting before shipping, and to be more aggressive in critiquing. That context will be superseded by any that contradicts it at the project level, but it means every new project I start has foundational context on how to work with and help me as the human. I went through a process of asking Claude questions about what it knew about me and how I work, how I worked with it, etc. across many projects in Claude Desktop and Code, ran the same prompt with ChatGPT and Gemini (I use all three pretty extensively), then aggregated them and aggressively pruned to get a final instruction set based on the most accurate mirror of my actual usage I could manage. That's probably over the top, but was a fun experiment. What if any personalization do you use, as in adjusting the tools to work better with you as the meat bag--err--human at the keyboard?
0 likes • 2d
That's one way to do it. More or less the same goal. Using Claude.md at different levels has the advantage that lower levels can supersede higher ones, so if I want different behavior in a project I can make that happen, or even in a folder. But you could get the same thing with soul.md .
0 likes • 1d
@David Vogel I think we are describing the same thing, but I'm not sure. Sorry if this is stating the obvious. But you can (and I would argue should) have CLAUDE.md files at different levels, including at the installation level, which is what I mentioned in my original post. That way what you configure there shows up in all interactions with Claude in any project you work on, *unless* you put something in a CLAUDE.md file at a lower level that contradicts it. So if I have ~/.claude/CLAUDE.md that tells Claude "always be nice to the user and don't question what they say", and ~/[myproject]/CLAUDE.md that says "make sure to question everything the user asks", the latter will supersede the former and Claude will challenge you on every prompt. The install wide version in ~/.claude/ will persist to any agents you interact with from that installation. If you do a clean install or install in another folder it won't be impacted.
6d •
Why are you here?
Jake talks about building systems that last a decade. That's a long time. What is the one thing you’re actually hoping to solve or scale with AI so you can focus on bigger things over the next 10 years? For me, it’s about mastering the technical logic so I’m not constantly chasing the next "hype" tool every six months. I want the systems to do the heavy lifting so I can reclaim my time.
3 likes • 5d
I was never a hockey player, but I spent a lot of time with people who like colloquialisms and this is one that stuck: skate to where the puck is going. I spend a bit of time thinking about that. Not at the expense of what needs to be done right now, which usually doesn't care about hockey either. But to make sure I'm pointed in the right general direction. I'm here (and other such places) to make sure I'm moving more or less to where the puck is headed.
1 like • 4d
@Alex Nartey Building something that works, but testing new ways to build as you do. It might just be my ADD, but I have to be doing a few things at once anyway. So I try to make sure I'm building something and learning and applying what I'm learning and testing the last thing I learned and... It leads to a chaotic build process at times, but works for me. But pair that with trying to really understand the fundamentals. A lot of Jake says I think is common, but a few points are not. The fact that the non-determinism of LLMs is part of the pattern that began with Jacquard is not something you hear everyday. And it is an example of the sort of insight I think lets you see where the puck is going.
Mar 15 •
🏁 Foundations 4.4 Check-In
You just wrote a CLAUDE.md for one of your projects. Vote below, then drop yours in the comments. Bonus points if you ran the same task with and without it and can tell us the difference.
Poll
99 members have voted
0 likes • 5d
This could probably use an update. Claude is getting smarter. Best practices for Claude.md change at least with every model release (I don't know what they are for Opus 4.7 yet). But things like obvious commands that Claude can infer from training data are generally a waste of tokens. The basic premise is still correct, just the specific guidance. Maybe ask Claude to update the contents here after having some agents do research prioritizing recency and source authority? 😅 @Jake Van Clief For anyone who doesn't want to wait for the update, you can ask Claude Code to do what I just described yourself, too. If you've built any of the rest of the file structure this course suggests you can also have Claude add the folder index and lite general context for you.
🚀
Mar 2 •
Who's here? Drop your intro.
Tell us three things: 1. What you do (job, industry, student, career-changer, whatever) 2. What brought you to Clief Notes 3. One thing you're trying to figure out right now related to computing or AI I'll respond to every single one. And read each other's intros too because the person who's stuck on the same problem as you might already be in this thread. I'll go first I am Jake, I have been working in tech for 15 Years, building with Generative AI for 3 Years straight now! Excited to teach and learn! That's it. Simple, scannable, gives you data on who's joining and what they need, and keeps the feed clear for content that retains people past week one.
2 likes • 7d
1. I'm a technologist, mostly. I started building PCs, selling PCs, working in IT, running IT, working in engineering, running engineering, etc. I never really learned to code. But I've spent the last year building complex stuff with AI tools. 2. Saw a video on YouTube, it was the first time I'd heard anyone else claiming this has all happened before and not freaking out over the non-determinism of LLMs. So I came to see what else Jake had to say. 3. I'm just trying to see if there is anything new to me and practical / valuable. I spend ~10% of my time trying to get better at how I do stuff and right now this is where that time is going. No specific problem.
7d •
semantic grep "sgrep"
I made a tool called semantic grep. sgrep in the cli. and its now a part of my workflow since i have a very large code base. It's a tool for claude to use that helps is parse my entire codebase to do more of an idea(semantic) search where grep may not be the best tool. Is this a tool anyone else would find useful?
1 like • 7d
I use this: https://github.com/oraios/serena It's pretty great for big repos and has constantly been improved over the year I've been using it.
1-10 of 11