Josh Jeong

Father. Branding/Marketing

[ℏεsam @Hesamation]

LOCAL LLM GUIDE (June 2026)

Cheapest full build: 1× used RTX 3090 (24GB) + rest of PC ≈ $1000-1500
16GB all-rounder → Gemma 4-12B
32GB all-rounder → Qwen3.6-27B
Agents & tool use → Qwen3.6-27B
Deep reasoning → Nex-N2-Mini

[Amber Shen @whosamberella]

I didn't touch TouchDesigner myself. Hermes agent learned it from scratch and built this:

→ navigated my desktop with computer use
→ figured out how to connect to TouchDesigner
→ read my reference images
→ iterated on the art with me in a self-learning loop
→ then saved what it learned as a reusable skill for the next image

all powered by @MiniMax_AI M3 × Hermes Desktop Agent @NousResearch

Here's a full breakdown 📽️

[ClaudeDevs @ClaudeDevs]

Claude Managed Agents can operate in a sandbox you control, on your own infrastructure or with any provider you choose.

Today we added new guides for @blaxelAI, @e2b, @googlecloud, @namespacelabs, and @superserve_ai, so you can choose the best fit for your use case.

[Pavel Durov @durov]

We now support rich formatting for all chatbots.

Tables, nested lists, inline media, formulas, headers and more — right in Telegram messages.

🔨 Start building! Docs: https://t.co/zgzPOOUJF5

[Ryan Carson @ryancarson]

Interesting

[Ziwen @ziwenxu_]

WE ARE SOOOOO BACK!

Someone leaked the Claude Fable 5 system prompt and ran it on an Opus 4.8

Output is like 90% of the real thing

Turns out half the magic was never the weights. It was the prompt the whole time

Repo down below:

[Sharbel @sharbel]

9 GitHub repos this week that make your AI agents actually useful:

1. last30days-skill - researches any topic across Reddit, X, YouTube, HN, and Polymarket automatically
https://t.co/jSgEaixGd9

2. headroom - compresses logs and files before they hit the LLM, 60-95% fewer tokens
https://t.co/pXgRjW9GJV

3. pm-skills - 100+ agentic skills for PMs, from discovery to launch
https://t.co/c2eHs1Hc6g

4. apple/container - run Linux containers as lightweight VMs on Apple silicon
https://t.co/KkGH36pdrk

5. Agent-Reach - gives your agent access to Twitter, Reddit, YouTube, GitHub with zero API fees
https://t.co/Bwjgn68nbE

6. open-notebook - open source NotebookLM with more features and flexibility
https://t.co/rx4oOX1Vmr

7. taste-skill - stops your AI from generating generic, boring outputs
https://t.co/9Mkx6J8fQU

8. MarkItDown - converts any file or Office doc to Markdown instantly
https://t.co/FXgdnGLeDR

9. NVIDIA Cosmos - open platform of world models for robots and autonomous vehicles
https://t.co/xDuAyeLKSS

Bookmark this and send it to your AI agent.

[0xSero @0xSero]

We got Mythos at home.

[0xMarioNawfal @RoundtableSpace]

A DEVELOPER JUST OPEN-SOURCED AN AI SHORTS FACTORY

• Generates TikTok, Reels, and YouTube Shorts automatically
• Free, open-source, and already has 13K+ GitHub stars

Repo: https://t.co/VBmtIloPyw

[Nous Research @NousResearch]

Introducing the Hermes Agent Profile Builder

You can now build a complete profile in the dashboard with full control over identity/name/description, model/provider, built-in + optional skills, skills-hub installs, and MCP servers in one easy flow

[Garry Tan @garrytan]

Nessie just became the best way to get all your existing context, memory and history from ChatGPT, Perplexity, and Gemini into all the other places you have memory, and also get it into OpenClaw/Hermes Agent. Their OpenClaw and MCP servers are ace.

https://t.co/0L1JaBlTvg

[SIMANTO @AIWithPRO]

THIS IS INSAAAAANE! 🤯

Google Gemini can now edit your videos and create amazing content better than CapCut in seconds

Use these prompts right now👇

[0xMarioNawfal @RoundtableSpace]

THE GUY WHO WON ANTHROPIC’S HACKATHON JUST OPEN-SOURCED HIS ENTIRE CLAUDE CODE SYSTEM

[kaize @0x_kaize]

120 AI models. FREE for a year. no credit card.

Hermes Studio already has NVIDIA preset as the base_url - so all you need is a FREE API key and you're running 120+ models instantly

here's the setup:

1/ go to https://t.co/vKcoAoknWf
2/ register, log in, bind your phone number
3/ grab your API key
4/ drop it into Hermes Studio

what you get:

1/ 120+ models
2/ 40 requests per minute
3/ free for a full year

while everyone's paying for API access, this is sitting right there for free

[sopersone @sopersone]

THIS GUY SHOWS THE FULL HERMES SETUP THAT TURNS A NORMAL LAPTOP INTO A $0 LOCAL RESEARCH AGENT IN 30 MINUTES

inputs: product idea, competitors, review pages, pricing tables, and client instructions

outputs: market reports, research workflows, saved skills, and tasks that get faster after each run

the weird part is not that Hermes can write a report. the weird part is that it keeps the method after the report is done

one task becomes a markdown skill. ten tasks become a small operating system for research, outreach, summaries, and client work

most AI tools reset the moment you close the tab. Hermes starts building memory on your own machine

that is why the setup matters. it is not another app to open. it is a local worker that slowly turns repeated work into reusable infrastructure

[Julian Goldie SEO @JulianGoldieSEO]

𝗚𝗲𝗺𝗺𝗮 𝟰 𝗳𝗿𝗼𝗺 𝗚𝗼𝗼𝗴𝗹𝗲 𝗷𝘂𝘀𝘁 𝗱𝗿𝗼𝗽𝗽𝗲𝗱, 𝗮 𝗳𝗿𝗲𝗲 𝗮𝗴𝗲𝗻𝘁𝗶𝗰 𝗺𝗼𝗱𝗲𝗹 𝘆𝗼𝘂 𝗰𝗮𝗻 𝗿𝘂𝗻 𝗿𝗶𝗴𝗵𝘁 𝗼𝗻 𝘆𝗼𝘂𝗿 𝗹𝗮𝗽𝘁𝗼𝗽.

It's a lightweight model that fits on a normal 16GB laptop.

But it reasons near the level of models twice its size.

Every AI agent needs a brain, and that brain used to live in the cloud.

So your data left your machine for every thought.

Gemma 4 puts a little AI right on your laptop instead.

It reads words, pictures, and audio all in one brain.

Plug it into Hermes and it builds games, tools, and animations.

No Wi-Fi on a plane? It still works, private and offline.

It's free and unlimited, and nobody can switch it off.

Want the setup? DM me.

[shmidt @shmidtqq]

NVIDIA just made paying for AI feel optional.

Open model, a million tokens of context, free tier with no per-token cost, runs on your own hardware.

Entire codebases, whole data rooms, a year of chat logs, all swallowed in one prompt. No chunking, no RAG, no rate limit theater.

The closed-AI premium has 90 days to defend itself.

Bookmark this and come back.

Open beat closed. Again.

[0xMarioNawfal @RoundtableSpace]

GOOGLE JUST FOUND A WAY TO SHRINK 31GB OF AI MEMORY DOWN TO 4GB

[Gipp 🦅 @gippp69]

A CHINESE STUDENT USED KIMI K2.6 TO RUN 300 AGENTS AT ONCE AND TURNED A $4,800 RESEARCH JOB INTO A 3-HOUR REPORT

kimi doesn’t work like a normal chatbot. he gave it one goal, and it split the task across hundreds of parallel agents before stitching everything back into real files

one agent checks reddit complaints. another scans tiktok angles. another maps app store reviews. another compares pricing, features, gaps, competitors and user pain points

the output is not just a long answer in chat. it can return reports, spreadsheets, dashboards and presentations that look like something a small research team would charge $3,000-$6,000 for

the scary part is the speed. what used to take 10-14 days manually can now be compressed into one prompt, a few hours, and a stack of files ready to use

most people will still use ai to ask basic questions. the edge is using systems like kimi to run the work in parallel while you only choose the question and make the final decision

300 agents won’t make a bad idea good. but they can show you the market, the gaps, the complaints and the angles before you waste 3 months building blind

[Machina @EXM7777]

this agent might perform better than OpenClaw and Hermes, just because of how it's built...

most agent frameworks are one giant pile of code where everything is wired into everything else

they're fast to start, but the moment it grows, one change breaks the whole thing and it takes a full day to fix it

OpenSquilla did quite the opposite:
> the core is tiny (around 100 lines)
> its only job is to decide what happens and hand the work off
> the model, the memory, the tools all live outside it as separate pieces
> you can swap or replace anything without touching the core

but the part that actually got my attention is what sits on top: MetaSkill 3.0

it solves a VERY important problem

your agent gets better, so you give it more skills and tools... which is fine

but the moment you ask it to chain a bunch of them into a real workflow, the combinations break and someone has to sit there hand-wiring the steps and writing constraints

which kind of defeats the point of having an AI do it for you

their MetaSkill flips that:
> it's a protocol that tells the model how to find the right skills
> pick the ones that fit
> and compose them into a working sequence on its own

you describe the goal in plain language, it organizes the execution behind the scenes

there's also MetaSkill that writes new MetaSkills:
> you hand it a request
> it drafts a brand new reusable workflow
> runs it through quality checks, and saves it

so the agent isn't just running skills anymore... it's organizing its own

very interesting approach, started building on top of this architecture

#SayItBuildIt