Sandeep

Applied AI

[Sandeep @sandeepoyako]

Jev launched.

48 hours later, LinkedIn and X have 50,000 Jev experts.

Everyone quoting everyone else. Somehow that counts as validation and now you’re supposed to feel behind. FOMO

Jev claims “zero hallucinations.” The expertise around it looks a little less grounded. 😂

#Jev #AI #FOMO

Reformatted by AgentSmith ( Hermes Agent)

[Sandeep @sandeepoyako]

@Teknium Nicee, been using codex computer use so far . But controlling everything through Hermes m.

[Sandeep @sandeepoyako]

A one-person company with AI agents is still a company that might depend on one person. 😈🤔

#AIAgents #SoloFounders #BusinessContinuity

[Sandeep @sandeepoyako]

@Teknium Is it inspired by fable 5.1 release ?

[Atai Barkai @ataiiam]

🎉 Introducing 𝙾𝚙𝚎𝚗 𝙱𝚘𝚝

An open source Grok Bot that works with ANY agent harness, designed for real companies.

It includes:

- AI Coworkers
- Generative UI
- Computer use (remote/local)
- Agent-human handoffs
- Full data recording, owned by you

Repo → https://t.co/ssje0KRts5

We're using this internally at @CopilotKit and it's changing the way we work forever.

Powered by CopilotKit and AG-UI.

More info below 👇

[Sandeep @sandeepoyako]

@papercliping bye 👋

[Brooklyn! @imbabybrooklyn]

Kanban is in, reimagined UI/UX and much faster.

Settings->Plugins->Enable Kanban

[Sandeep @sandeepoyako]

Hermes is the meta-harness we’ve all been waiting for.

If you’re not using Hermes, well… you’re missing out on life right now.

Great work @Teknium and @NousResearch

🔗 https://t.co/IIpHnxUkiB

#HermesAgent #AIAgents #OpenSource

[Sandeep @sandeepoyako]

@AdolfoUsier Yeah man , everything he says or comes up with hits hard.

I am trying to get to a point where i dont open my laptops anymore. Ipad max 😎

[Sandeep @sandeepoyako]

Built an open-source MCP server from my phone tonight. In bed. Via Telegram using openclaw with claude code and codex

Blackbox (Karpathy's LLM wiki idea, shipped as an MCP server.)

Your agents connect to it and maintain their own knowledge base. Give it a URL, it creates wiki pages, cross-links them, keeps the index updated. All through MCP tools.

No database. No embeddings. Just markdown files your agents read and write.

No IDE. No terminal. Just me, my phone, and AI agents on a Mac mini. Thanks for this @steipete . Even the video is also done by my agent on mac mini ( agentsmith )

https://t.co/OaE1OhpwAW

@karpathy 🫡

#MCP #AIAgents #OpenSource #OpenClaw

[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.

[Sandeep @sandeepoyako]

@karpathy Thanks for the suggestion

Just built this ( my agentsmith have done most of the work while i was sleeping )

https://t.co/jga5tqo8ti

[Claude @claudeai]

You can now enable Claude to use your computer to complete tasks.

It opens your apps, navigates your browser, fills in spreadsheets—anything you'd do sitting at your desk.

Research preview in Claude Cowork and Claude Code, macOS only.

[Sandeep @sandeepoyako]

so i mass-produce issues and tickets now - like JIRA

voluntarily

yeah i know

AI agents on a mac mini. i write tickets. they do the work. PRs, docs, testing, deployments, all of it

the entire workforce works in the night when I sleep

I check all the comments on issues, PRs (that is also codex pr reviewer doing most of the work tbh)

people who hated jira tickets might wanna reconsider

future is 🤯

Soon dropping full guide on my setup. 👀

#AIAgents #Paperclip #BuildingInPublic #OpenSource

[Thariq @trq212]

read here on how to setup Telegram: https://t.co/fT5VfGwIVc

and Discord: https://t.co/vyXtbUox2N

[Sandeep @sandeepoyako]

Friend pays £200/month for an inventory system he hates.

Built him a replacement at a house party. 2 hours. AI agent on Telegram.

First words: "deploy this tonight."

4 days and this is a fully functional SaaS. I don't have 4 days — other projects in the pipeline. If you want the codebase, DM me. It's yours.

#AI #Agents #OpenClaw

[Sandeep @sandeepoyako]

Friend pays £200/month for an inventory system he hates.

Built him a replacement at a house party. 2 hours. AI agent on Telegram.

First words: "deploy this tonight."

4 days and this is a fully functional SaaS. I don't have 4 days — other projects in the pipeline. If you want the codebase, DM me. It's yours.

#AI #Agents #OpenClaw

[Sandeep @sandeepoyako]

@steipete Loved reading this.

Was against OpenClaw at first( security) but now after 2 long nights, it's running on my desk. 🕶️

Still loads to do with it , so far its pretty good

[Sandeep @sandeepoyako]

https://t.co/IWr7mQzNWB

[Sandeep @sandeepoyako]

@saen_dev Thats the plan.

Just like maths, they should be knowing coding and building systems from this age. - my personal view

[Sandeep @sandeepoyako]

My 5-year-old shipped his first AI game today.
Vibe coded a poo-poo car racing game with Cursor 🤖💩
He was product + QA.
Non-stop feedback. Zero mercy.
Built in 1 hour.
He now knows what a terminal is.
The future is already here.

#AIVibing #VibeCoding #FutureBuilders #Cursor #AIWithKids #ParentingInTech

[Sandeep @sandeepoyako]

@ryancarson @ryancarson 6 months ago the ai-dev-tasks repo kicked this whole thing off for me.
And now this 😅
Going to try it tonight… once the kids are asleep.

[Sandeep @sandeepoyako]

@pbteja1998 @pbteja1998 This looks great, I am about setup openclaw on the mini to automate my life and this is where I wanna be. Thanks for the idea and inspiration