fullstackchris

Code design? Perhaps. But how are you going to inform a model of every sprint meeting, standup, decision, commit, feature, and spec that is part of an existing product? It's no longer a problem of intelligence or correctness, its a problem of context, and I DON'T mean context window. Imagine onboarding your companies best programmer to a new project - even they will have dozens of questions and need at least a week to make productive input to the project. Even then, they are working with a markedly smaller scope of what the whole project is. How is this process translatable to an LLM? I'm not sure.

> but about putting some requirements stated in a slightly vague fashion, and actualizing it for the stakeholder to view and review

sounds like a form of creativity to me!

Good thing we're all working here in Switzerland :/

Still cracking away CodeVideo, a way to create software educational content in a declarative manner. Design your course once, export it to every format you can think of (video, PDF, markdown, HTML, and more). I was recently inspired by the feature set I saw at Scrimba, so we just added slide functionality! The blog post is here: https://codevideo.substack.com/p/introducing-the-slide-displ...

And an example video is here: https://youtu.be/1duE604MGHs

It definitely has not gotten the traction I'd expected, but at the very least I'm very close to start making my own courses with it!

Very impressive and cool stuff! This shares similar themes to what I'm building over at https://studio.codevideo.io - I also chose to go with an event sourcing solution. I experimented with real time mouse recording but though it was overall too complex for more advanced examples, so for now I've setted with a virtual mouse that you can move from place to place to simulate how it might actually move - I'm curious, how did you solve realtime edits and highlighting directly in the editor? Are you using Monaco editor with realtime highlighting updates somehow? It's something I still haven't implemented in CodeVideo...

If I see another mention of the paper clip example I'm gonna lose it.

Perhaps better is to kindly refer everyone to a physics 101 text book.

seen a lot with uvx as well (apparently a package manager for python, but i try to stay as far away as possible from python)

Not OP but perhaps the following mcp tools: google maps api, nearest supermarkets, puppeteer their product listing pages?

Though to be honest not sure why you would need so much info - if I need lettuce or tomatoes for example, I know theyre gonna be at essentially every supermarket in my area....

For me it was implementing a simple `execute_terminal_command` tool along with hooking up to my company's jira and gitlab (dont worry security gurus, for the command line, I have a hardcoded list of allowed read-only commands that the LLM can execute, and both jira and gitlab servers likewise have readonly options.)

What I will say is I agree there should be an option to get rid of the chat confirmations of every single new tool call in a chat - as well as building a set of "profiles" of different tools depending what I'm working on. Also strongly agree there needs to be an internal prompt possibility to explicitely tell the LLM what tool(s) to favor and how to use them (even in addition to the descriptions / schemas of the tools themselves) I opened an issue on the anthropic repo exactly about this: https://github.com/modelcontextprotocol/typescript-sdk/issue...

Moravec's paradox

This is super insightful, and I think there is at least part of what you are thinking of: an abstract syntax tree! Or at the very least one could include metadata about the token under scrutiny (similar to how most editors can show you git blame / number of references / number of tests passing in the current code you are looking at...)

It makes me think about things like... "what if we also provided not just the source code, but the abstract syntax tree or dependency graph", or at least the related nodes relevant to what code the LLM wants to change. In this way, you potentially have the true "full" context of the code, across all files / packages / whatever.

Thats what I'd like to know, is this the kind of process that we can _get_ to 6 sigma by more observation time? Or would we need other observations / thats "as good as it gets" for the Webb's capabilities?

Many trolls hiding in these comments... either that or they just dont understand what the scientific method is.

Bostrom's trilemma and the notion of the Great Filter!

But in the grand scheme of things, even its "bad news" just ONE datapoint of life elsewhere is at least something to start working with.

Love it - and ironically this is something one would struggle to build with "vibe coding" alone

and the next 10 after that, and the next 10 after that, and...

I write _extensively_: on my own personal blog [1], on my companies blog, on various substacks, on medium, and on dev.to... idk, what this author seems to suggest might happen with good writing and sharing hasnt really panned out for me... but who knows? Maybe I just suck at writing!

[1] https://chrisfrew.in shameless plug

Definitely agree there. There seems like there is a label for EVERYTHING these days. Coincidentally there was an article in the paper I saw just today about how destructive self diagnosis can be (and how many gen z / alpha are doing that via tiktok) I rememeber when some first tried to explain ADHD to me... and I was like... "what? elementary school kids dont like sitting in a classroom for 6 hours a day? so like, being NORMAL?"

Source: I had / (have?) ADHD, don't think about it at all anymore, just keep working hard, stick with hobbies and you can do great things regardless of what "labels" society has assigned you.

There is a $499 a month enterprise subscription hidden deep within the account page... ;) but what do you mean exactly? multiple accounts / live sharing sessions? Tied to an org's codebase?

> CodeMic aims to replace books and blogs for discussions of large codebases and showing changes to code over time

This is huge and something I've pondered about how to do efficiently, because what if you could forward this essential context info to something like an LLM, or even LLM / AI aside, address immediately the classic "why did we do that here, in this way, at this time?" questions about a codebase.

I could definitely see some CodeVideo / CodeMic interoperability via some sort of export :D

Let's see what happens...

Great catch - real link shipping to prod as we speak. Thanks for checking out the site!

:D lol well at least we know this wasn't generated by an LLM, that's for sure (sorry everyone, my friend's spam post)

Hey thanks for stopping by and the kind words! When I saw your launch post a few weeks ago (months ago? time has become irrelevant) here on HN, my heart skipped a beat, I thought CodeMic was direct competition to CodeVideo - maybe they are, or maybe they are different enough to co-exist :) I anyway immediately dropped my email for the early access!

For a brief moment, I looked into a sort of VS Code plugin for CodeVideo but I realized it would be tough with things like mouse and animating typing directly in VS Code (unless I did a complete fork like the Cursor path). But I wasn't quite ready to take all of that on.

But yes, I think it's a huge latent space that hasn't been addressed anywhere in the market simply because you need to go half insane to take something on as complex as replayable state management for the IDE.

Yes, the in-browser playback is quite finicky at the moment... for now the simple step-by-step and time travel buttons (next / previous) are safer...

Playback seems to be working on Chrome.. but I will make a note however for Firefox. Thanks for checking it out!

Ah yes... I know the pain of creating / editing / scrubbing through educational software courses... part of why I built this :) Thanks a lot for the kind words! And about the link to the video - good point! here it is:

https://coffee-app.sfo2.cdn.digitaloceanspaces.com/codevideo...

Added! Also made the more advanced action buttons (keyboard / spaces) collapsed / expandable.

Americans perplex me (and I am one) we live in the most capitalist focused economy in the world and yet everyone acts that they are guaranteed a job. Sorry, but even in "socialist countries" in Europe / rest of world this is not even the case. Then the country turns around and elects someone like trump - thinking collectively someone like that will improve this situation?

Especially the FAANG folks - they get pissed when getting fired from there like DUDE you are literally in the top 0.01 of earners on the planet get over yourself. Poor financial planning and overleveraging yourself are not excuses when you get laid off / fired!!!

Just go read stuff on Blind for more of what I am talking about... i.e "got laid off... nervous to retire with 7 million?!?" like hey, some of us here will probably never see that kind of money in our entire lives

What about those two vs. the concept of "software engineering"? - there, the "code" or "program" is even _less_ important, just a tool in an ecosystem where you are asking bigger questions like "is this maintainable / testable / readable / etc.?" "is this the right language / framework for my org / others to use" and so on and so on. These questions and context quite literally represent billions of context tokens in LLM world and why I continually do not worry about 99% of the fear mongering of them replacing anybody who writes code.

You seem like you know what you are talking about... mind if I ask what your thoughts on quantization are? Its unclear to me if quantization affects quality... I feel like I've heard yes and no arguments

> You've experienced so many startup failures that your LinkedIn profile should just read 'Professional Titanic Passenger: Always Picks the Wrong Ship'.

:'(