Vic

enjoying the internets

[Vic @vicgrss]

@guillaumemarq Exciting Times

[Vic @vicgrss]

@veherkeus @samuSCB15 @Paris ?

[Vic @vicgrss]

@samuSCB15 C’est le bar l’atmosphère qu’ils saccagent ?

[Vic @vicgrss]

@karpathy Why @obsdmd rocks : https://t.co/ioIzqEEFdA cc @kepano

[Andrej Karpathy @karpathy]

LLM Knowledge Bases

Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So:

Data ingest:
I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them.

IDE:
I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides).

Q&A:
Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale.

Output:
Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base.

Linting:
I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into.

Extra tools:
I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries.

Further explorations:
As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows.

TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

[Vic @vicgrss]

@byAnhtho congrats !

[Vic @vicgrss]

@NotionHQ when pressing search (command + K), autofocus doesn't work. could you fix it back ? Thx !

[Vic @vicgrss]

@pontusab @bnj @benhylak You ve stolen everything from @getbasalt !!!

[Vic @vicgrss]

Nice reading from @ttunguz - AI is better on raw text

useful lib to transform Document to text to AI
https://t.co/s5zZV06Blk

[Vic @vicgrss]

@Paris c’est top cette histoire de trilib ! Je ne comprends pas par contre pourquoi je reçois le prospectus x3 dans la boîte aux lettres !bb

[Caleb Watney @calebwatney]

This is the best paper written so far about the impact of AI on scientific discovery

[Vic @vicgrss]

@OUIGO 7622 c’est l’anniversaire de ma femme si vous pouvez faire un mot au micro c’est sympa. Elle s’appelle Blandine 👋

[Vic @vicgrss]

why email that self-destruct don't exist ? spare some storage by default

[Vic @vicgrss]

I love that https://t.co/XKB0KMx5hE

[Massimo @Rainmaker1973]

Zooming into an iPhone CPU die (5nm)

⁽ᵛᶦᵈᵉᵒ ᵉᵈᶦᵗᵉᵈ ᶠᵒʳ ᵉᵈᵘᶜᵃᵗᶦᵒⁿᵃˡ ᵖᵘʳᵖᵒˢᵉˢ⁾

[scott belsky @scottbelsky]

Collapse the talent stack every chance you get.

As I reflect on the teams I’ve led and hundreds of start-ups I’ve worked with, there is a consistent unfair competitive advantage i’ve witnessed when the talent stack was collapsed - when the lead designer was also the product leader, when the front-end engineer was also a designer, when the designer is also a great copywriter, when the product leader was also the founder/ceo, etc. Tighter conduits for decision making and synthesizing information are an incredible advantage when it comes to crafting products. Many start-ups enjoy the benefits of collapsed talent stacks and then undo them as they grow (and most big companies just don’t understand this). In your hiring (and your consolidating), I encourage you to collapse the stack whenever you can. Especially given all the focus on “product led growth,” these days (which really means helping new customers feel successful more quickly, discover the benefit of sharing, people talking about the product doing things they didn’t expect), all of these are as much marketing driven experiments as they are “traditional product specs” and design explorations! Collapse the stack. While it might feel like “double duty” to your leaders, it works magic - especially in early stage products or periods of self-disruption where you need to speed up exploration and execution.

I go into detail on this and share a bunch of other Implications for the way modern teams should be structured on Implications. (implications. com)

[Vic @vicgrss]

https://t.co/8IfVBZpTdm est live sur @ProductHunt !
Your team’s wireframing tool - right from Figma, enjoy and support !

[Vic @vicgrss]

I have a feature request! @NotionHQ, please let me play with database fields in the page. Will you close the feedback loop with me? @CycleProduct might be the easiest way to do it --> https://t.co/xQ9PiMq53G

[Vic @vicgrss]

https://t.co/qrj5NpLzSW