Jashan Preet Singh

9:00 PM. 200+ pitch decks. Zero bandwidth. Every VC knows this pain: 👉 Too much deal flow 👉 Not enough signal 👉 Fear of missing the next unicorn What if your inbox filtered itself? beeeowl = your #privateAI infra🦉 Book: beeeowl.com #VentureCapital #VC

@karpathy basically @NotebookLM . right?

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.

Every time you hit “Send” on a cloud AI prompt, you’re making a trade: convenience for control. Your proprietary data? Processed on someone else’s servers. That’s why we built beeeowl.com 🦉 100% secured, private #AI environments. Pre-configured. Zero headaches.

Imagine a brilliant digital employee running on a dedicated Mac Mini right on your desk—100% private, military-grade security, and zero external eavesdropping. 🚀 Get your private AI live in under a week: beeeowl.com #PrivateAI #SovereignAI #OpenClaw #Beeeowl

AI makes execution free. So what's left? Verification, trust, and liability. When AI can produce 100 versions of anything, the value goes to whoever says "I'll stand behind this one." Your work is not your output anymore. It's your judgment.

@shannholmberg share the outputs.

claude superpowers is the most underrated plugin for marketers right now 83,000 github stars. trending daily. but almost everyone using it is a developer here´s how it works and how to apply it to marketing 🧵

@indie_daniel @waynesutton @convex @supabase good to know.

@singhjp006 @waynesutton @convex @supabase Missing from the list: Convex also comes with a built-in cache that invalidates automatically when the underlying data changes. Plus things like workflows, caching of external data, not having to manage a growing list of migration files etc. It's a more complete backend.

Let's settle this down. Have lately seen a lot of people talking about @convex, and almost everyone is talking about how convex is another gimmick and not at all scalable, etc. I have used @supabase , I am comfortable with SQL, psql, etc. What's the hype about? I understand some out of the box things, but apart from that, is there anyone who can vouch for convex and running a scaled up production application except @theo 's t3.chat?

@mikefutia @cluely SLIDES

So @cluely is getting 400k+ views with AI TikTok slideshows. And I just automated the entire system 🤯 AI selfie hook → listicle slides → soft CTA on the last slide. Rinse. Repeat. Scale. And now my n8n + Airtable system runs it on complete autopilot. Perfect for e-comm brands & SaaS companies who want to test this format without the manual grind. Here's why this format is blowing up: → AI-generated "selfie" as the hook image → 5-8 slides with tips, hacks, or advice → Text overlays generated automatically → Promotes product on the final slide → Music added automatically on publish 362k views. 462k views. 492k views. Same playbook, multiple accounts, fully automated. Here's what my system does: → Define your niche, tone, and content style in Airtable → AI generates slide text and hooks automatically → Nano Banana creates every image with text overlays → Preview and approve before anything goes live → Blotato publishes to TikTok on your schedule—with music No manual Canva work. No writing slides one by one. No posting by hand. What you control in Airtable: * Niche and content direction * Hook style and tone * Image aesthetic and reference images * Posting schedule (days + times) * Approval before publish Set it up once. Let it run daily. I recorded a full step-by-step tutorial showing exactly how to build this system. Want access for free? > Like this post > Comment "SLIDES" And I'll send it over (must be following so I can DM)

@kiwicopple @waynesutton @convex @supabase as in a separate environment from prod which comes at no cost with convex. supabase do offer that, but in the paid tier as Branching.

@singhjp006 @waynesutton @convex @supabase Genuine question: what does a live dev environment mean?

@theo @smotched @convex @supabase I got your point. I do not have a specific use case as such, just looking to try convex for my next SaaS product after watching your video. But I misunderstood how Edge works which is why this question. Thanks for clarifying. It helps.

What is your use case? The whole industry has largely moved away from edge functions, since round trips to your database are more frequent than round trips to your users. Edge funcs are best used for things cached at edge, like HTML assets you'd put on a CDN. Convex runs your functions on the same box that the DB lives. That makes them significantly more performant. Afaik, Supabase doesn't run their functions themselves at all, so actually accessing the data will be significantly slower for no good reason. Here's a video that might help address your misconceptions about how server runtimes work youtube.com/watch?v=lAGE-k…

@smotched @theo @convex @supabase yes.

Agreed. Can you please answer just one question (out of a lot of I have): Isn't Convex not having edge functions a huge gap? If you think, no, I want your understanding about it. Quick google summary: Cloud functions run in centralized data centers, while edge functions are deployed on a global network of servers (the "edge") and execute physically closer to the end-user. This fundamental difference leads to distinct performance characteristics and use cases.

@singhjp006 @smotched @convex @supabase You have such Redditor energy man 1. Not everything is a debate 2. You have to make points in order to have a debate 3. You still haven't even tried it. Just try it. Your questions will be answered quickly when you do 4. I have covered the downsides extensively in my vids.

@theo @smotched @convex @supabase Again, I was genuinely curious. But I don't think you have grown up enough to have a tech debate. Remember, no tech has ever been invented that is 100% good and will be relevant always. Suggestion for next time, try to do a healthy discussion. thanks,

@smotched @singhjp006 @convex @supabase I can't address points when there are no points to address.

@AriesTheCoder @theo @convex @supabase Agreed.

@theo @singhjp006 @convex @supabase Man, this insidiious practice of self high fiving needs to stop. @singhjp006 would've happily engaged with you to discuss this without you resoting to tactic. Still open by the way. You could create a new thread and you'll find that people will engage with you there too.✌️

@theo @convex @supabase you are trying too hard. chill buddy. I was being genuinely curious. but seems like someone took it personal.

@singhjp006 @convex @supabase Going to YouTube comments for tech suggestions is like going to McDonald’s to learn how to be a doctor

@theo @convex @supabase I expected a better reply from you @theo . Nevermind, I was fully sold on using convex, but then I read Reddit and comments on your videos. And this is where I wanted to have a public opinion from my network on x.

@singhjp006 @convex @supabase Never mind. I read your other comments. It’s clear that you really doing want to use convex. You’re pretending you “just want to ask questions” but your intent is malicious at best. Keep using what you want to use. You’ll grow up in a few years :)

@waynesutton @convex @supabase thanks for sharing those links. you still missed to answer my question related to conext not having edge functions but cloud functions.

@singhjp006 @convex @supabase Please take 15 min. 5 and watch this video youtu.be/hiptYyfndxc?si… and 10 to run through the tutorial. docs.convex.dev/tutorial/

@benktz @convex @supabase @theo @benktz what do you think about this? x.com/singhjp006/sta…

Theo's YT comment section is almost always a cess pool. Not even worth giving it a second thought. The best validation for @convex is to just try it. It takes <5 minutes to spin up a demo app and play around with it. The fact that I can write a query (IN TYPESCRIPT!), then call that exact query in a typesafe way from my frontend, without needing some annoying ORM bullshit, AND it automatically syncs in realtime....its just perfect. I also like that your schemas are 'declarative' by default too. I don't need to worry about migrations. With Supabase, my 'declarative' schema was never generated properly, I still had to manage a massive folder of migrations, and unless you use RPC a lot, you have to put a ton of logic on your own backend somewhere else. It's a beautiful abstraction that reduces my mental load and let's me just ship things.

@waynesutton, if you will, out of this comparison, out of the 3 ❌ in supabase, only live dev environment seems like a ❌ to me. and for convest not having edge functions I think is a huge gap. no? Also, I like database staying in code for agentic coding, which is one of the biggest power of @convex and tbh, I am looking for reasons to use convex, but I want strong ones.

@singhjp006 @convex @supabase Just wait, there are so many.

@waynesutton @convex @supabase I want to hear that only. just want some validation.