Hacker News: lummm

Show HN: Codecrush – AI Documentation Generator

lummm — Sat, 03 May 2025 16:16:28 +0000

Hi, sharing something I've been working on for a while. CodeCrush runs locally and points to an LLM of your choice - so self-hosted, OpenAI, whatever you're comfortable with. It uses the LLM to extract the information it needs to build a knowledge graph over a codebase, and leverages that knowledge graph to build documentation over important sections of the codebase it identifies. You can also do more targeted documentation generation by giving it a prompt.

Comments URL: https://news.ycombinator.com/item?id=43880039

Points: 2

# Comments: 0

Show HN: Codebase Knowledge Graph Builder

lummm — Tue, 22 Apr 2025 01:56:57 +0000

I've been working on tools for 'understanding' software code for a couple years now. I have yet to find a better way to model actual concepts than with knowledge graphs. To that end, I present to you here you a tool to automatically construct a knowledge graph over a codebase, which can serve as the foundation for a wide range of downstream tools. Point it at a codebase, give it your OpenAI API key, and let it go to work. Larger codebases (> 1000 files) will take a while to process - a conservative estimate would be 12 files / minute.

Comments URL: https://news.ycombinator.com/item?id=43758410

Points: 4

# Comments: 0

New comment by lummm in "Show HN: I built an AI that turns GitHub codebases into easy tutorials"

lummm — Tue, 22 Apr 2025 00:08:28 +0000

I actually have created something very similar here: https://github.com/Black-Tusk-Data/crushmycode, although with a greater focus on 'pulling apart' the codebase for onboarding. So many potential applications of the resultant knowledge graph.

New comment by lummm in "How productive will GitHub Copilot make me?"

lummm — Tue, 25 Feb 2025 03:38:22 +0000

This my most honest take on how effective Copilot can be if it's given an honest shake. Do you agree with my assessment?

How productive will GitHub Copilot make me?

lummm — Tue, 25 Feb 2025 03:38:22 +0000

Article URL: https://blacktuskdata.com/copilot-and-productivity.html

Comments URL: https://news.ycombinator.com/item?id=43167880

Points: 3

# Comments: 1

New comment by lummm in "Late Chunking and better RAG chunking"

lummm — Wed, 18 Dec 2024 15:44:31 +0000

Late chunking is a promising approach to computing text embeddings that incorporates more of the actual document into each chunk's embedding. I think it's pretty promising.

Late Chunking and better RAG chunking

lummm — Wed, 18 Dec 2024 15:44:30 +0000

Article URL: https://blacktuskdata.com/late_chunking.html

Comments URL: https://news.ycombinator.com/item?id=42451376

Points: 3

# Comments: 1

New comment by lummm in "A Barebones GraphRAG Implementation"

lummm — Wed, 27 Nov 2024 16:05:47 +0000

Mostly as an academic exercise, I built a much simpler implementation of the Microsoft GraphRAG implementation laid out in this paper: https://arxiv.org/abs/2404.16130. I find it to be nevertheless pretty functional, and certainly more hackable than the de-facto implementation.

A Barebones GraphRAG Implementation

lummm — Wed, 27 Nov 2024 16:05:47 +0000

Article URL: https://github.com/Black-Tusk-Data/minikg

Comments URL: https://news.ycombinator.com/item?id=42257102

Points: 2

# Comments: 1

New comment by lummm in "A Developer's Guide to Large Language Models"

lummm — Fri, 01 Nov 2024 16:23:19 +0000

This is a small book / guide I've written with the goal of demystifying the details of language models for developers without any ML / neural network experience. From my perspective, there is a level of understanding within the grasp of most developers that will enable them to build software with drastically new capabilities. This guide is my attempt at bridging that gap in education. Please share if you find it helpful!

A Developer's Guide to Large Language Models

lummm — Fri, 01 Nov 2024 16:23:19 +0000

Article URL: https://devs-ai-guide.blacktuskdata.com/

Comments URL: https://news.ycombinator.com/item?id=42018572

Points: 1

# Comments: 1

New comment by lummm in "The extensible vi layer for Emacs"

lummm — Fri, 03 Mar 2023 14:44:14 +0000

I've always had view-mode come on by default, and I have a bunch of custom keybindings for view mode which are nicer on the hands when you're just reading text.

So (add-hook 'find-file-hook 'start-view-mode) to turn it on automatically.

(defun view-mode-background () (if (bound-and-true-p view-mode) (face-remap-add-relative 'mode-line '((:background "#9400D3"))) (face-remap-add-relative 'mode-line '((:background "red"))))) ^ this helps a lot to know whether or not you're in view mode

And then: (defun view-mode-keybindings () (define-key view-mode-map (kbd "j") 'View-scroll-line-forward) .. etc

Ask HN: Where to engage with Vancouver tech scene?

lummm — Thu, 26 Jan 2023 16:57:12 +0000

I've been working in the Vancouver tech industry for 6 years this January, but have yet to feel as if I've 'connected' in any way with tech professionals outside of my coworkers. Is anyone aware of a canonical watering hole or really anything that brings together tech professionals in the city?

Comments URL: https://news.ycombinator.com/item?id=34534421

Points: 1

# Comments: 0

New comment by lummm in "Ask HN: What would be your stack if you are building an MVP today?"

lummm — Thu, 26 Jan 2023 16:48:00 +0000

I think I'd have to say NextJS with SQLite. Every DB table would just be 'id' and 'data', where 'data' is JSON-encoded and the schema is only enforced in code. This lets you write a 30-line ORM. Foreign-key relationships are just modelled as lists. If speed is the major concern, I find this the fastest way to build. You can bring structure to the DB as things are nailed down.

New comment by lummm in "Things your manager might not know (2021)"

lummm — Mon, 31 Oct 2022 16:00:16 +0000

What about GCP?

New comment by lummm in "Developers can't fix bad management (2020)"

lummm — Thu, 10 Jun 2021 14:07:20 +0000

I think below a certain threshold of competence, a manager will negatively impact the productivity of a team mostly by eroding any ability to focus on a single task at a time. A good manager should shield the team until it is appropriate from the very things a poor manager would be dumping on the team as they arise.

New comment by lummm in "Overwork contributes to 745k premature deaths per year"

lummm — Fri, 28 May 2021 20:08:00 +0000

I don't understand the lamentation over a culture of overwork. What other method is there to get ahead as an individual other than out-competing your peers? My understanding of life as a young professional in China or Korea is that the competition is almost unbelievable to someone from North America.

New comment by lummm in "[dead]"

lummm — Tue, 18 May 2021 02:05:49 +0000

So what are we supposed to do about it? Work less? Lol.

New comment by lummm in "TikTok Agreed to Buy More Than $800M in Cloud Services from Google"

lummm — Wed, 15 Jul 2020 13:51:11 +0000

I think the use of government surveillance is less likely to be used to conduct drone strikes and far more likely to prompt a Tiananmen square esque response to dissent. Thus I would firmly disagree with this sentiment.