Yannis

@RajputNikh13836 Both!

Mac mini or macbook

@sri9s Chief orchestration officer

They say AI will create entirely new jobs we can’t even imagine yet. So, name one

I've been using Hermes for a little over a week now. Around the same time, I started a new X account from zero to document the AI workflows I actually use. I expected the coding stuff to be the main use case, but the more useful part has been much more boring: small daily loops that leave me with files I can inspect before anything goes public. Here are 5 automations I added so far. 1. Daily X analytics Hermes pulls my profile and post data from the X API once a day: followers, posts, impressions, likes, bookmarks, reply ratio, best post, weakest post. Then it writes a markdown drop into my Obsidian vault and updates the performance notes. This is already useful because it tells me when I am lying to myself. One day I posted 17 replies and only 3 originals. Hermes showed me the whole day did around 224 impressions. Pretty obvious lesson: replies are useful for being in the conversation, but they are not a replacement for one strong post. 2. Daily AI news scan I use Firecrawl as the web scraper layer. It pulls pages from the AI tooling sources I care about, cleans them into text, and hands them to Hermes. Hermes then filters for things I can actually add an operator take to. If there is no workflow angle, no proof I can show, or no reason my audience should care, it gets parked. This matters because pure AI news is usually low-value unless you can explain what changes in the actual work. 3. Content that learns from previous posts This one is still early. The account is new, so I do not have much data yet. But the loop is there: Hermes reads what performed, what got bookmarked, what died, and updates the rules for future drafts. Right now the pattern is simple. Concrete Hermes + tool posts work. Abstract agent-memory posts do not. That sounds obvious after the fact, but having the system write it down stops me from repeating the same weak angle three days later. 4. A content pipeline with Notion + Obsidian Notion is my active idea board. Obsidian is the memory: voice notes, decisions, post-mortems, drafts, research, playbooks. Most days I still do manual research. I prefer it that way because the taste has to come from me. But when motivation is low, Hermes can look at the idea bank, the latest sources, and the performance notes, then give me drafts that are at least pointed in the right direction. The goal is not to remove the human part. It is to make consistency less dependent on having a perfect day. 5. Turning X posts into LinkedIn drafts I am not very active on LinkedIn yet, but I plan to be. So I added a simple workflow where Hermes can take a tweet or X article, read my voice notes, and turn it into a LinkedIn draft. It writes the draft to a folder first. No auto-posting. I still want to edit it, because LinkedIn punishes generic AI writing harder than X does. The thing I keep coming back to is that Hermes gets more interesting when it becomes the layer between tools. X API for analytics. Firecrawl for web research. Notion for planning. Obsidian for memory. Codex/OAuth/MCP for permissioned work. None of these loops are individually impressive. But together they make the account easier to run on days where I would normally do nothing.

@mynameisyahia I don’t think there is such a thing as an American name haha

I hope it’s not something like “if foreign name then kyc”

Most developers use Cursor like autocomplete. The interesting part is the stuff they never touch: > thermo-nuclear code reviews > cloud agents working while you sleep > reusable skills > rules that make every repo smarter I wrote up the full system.

@MehakdeepK81 Codex! I think the user experience is great and $100 gives you a lot of inferences.

Developers, you have $100. What are you buying ? -Claude -Codex

@hazzler @FredrikHjelm4 @gabriel1 Nope, I’m from France but I live in Sweden! What about you?

@yannisbuilds @FredrikHjelm4 @gabriel1 Yannis are you from Sweden?

Spot on from @gabriel1 (Swedish AI researcher/founder at the frontier). Countries like Sweden are AI users, not owners. If the US restricts frontier access, we could regress fast. Three paths: 1. Go all-in on sovereign models + compute (brutal but independent) 2. Stay dependent and negotiate special US/Western access 3. Find a smarter third way (alliances, open-source coalitions, compute partnerships?) What should small high-trust nations actually do?

@0xSero It’s coming right on time 🤝

State of Local AI #1 ——— In lieu of Fable ban. Here’s the best LLMs of the week to run on your hardware. —— 4-8gb vram/ram 500$ - Gemma-4-qat huggingface.co/unsloth/gemma-… I had someone mention it’s very good for subagent stuff —— 8-16gb vram/ram < 1k usd - Gemma-12B huggingface.co/google/gemma-4… without a doubt the smartest model of its size —— 16-32gb Apple/Strix halo 1-2k usd - Diffusion Gemma26B huggingface.co/google/diffusi… - on 1x 6000 it’s eating up to 600 tok/s - smallest smart MoE we have - lots of world knowledge - easy to run —— 32-96gb ram/vram (2-10k usd) - nex-n2-mini huggingface.co/nex-agi/Nex-N2… builds on qwen3.6-35B and seems to do really well - qwopus-27B huggingface.co/Jackrong/Qwopu… this model topped a lot of our benchmarks at local.ai —— 384gb vram (10-50K usd) - huggingface.co/MiniMaxAI/Mini… 23B means it’s close to qwen3.6-27B per token, while also have a lot of specialisation. - fast inference - top open weight model on AA —— 768gb-1TB - huggingface.co/moonshotai/Kim… Kimi has always been a top player here and their last model cuts speed and cost down by 30% - great vision support - first coder model by moonshot ——— Top models: 1. Qwen3.6-35B 2. Qwen3.6-27B 3. Step-3.7-Flash 4. Minimax-M3 5. Deepseek-v4-flash ——— Budget sweet spots: #1 - 1K usd Single 3090 / Mac mini / Intel arc b70 / AMD - Qwen / Gemma #2 - 5k usd DGX Spark / Mac m5 max / 4x 3090 - qwen / Gemma step and deepseek flash #3 - 12k usd RTX Pro 6000 / Mac Ultra / 2x Spark / 8x 3090 Ds4-flash / step-3.7-Flash and above #4 - 24k usd 2x 6000 / 2x Mac Ultra / 4x Spark / Mix Same as above #5 - 50k usd 4x 6000 / 4x Max Ultra / 12x Spark / 2 H100 Minimax-m3 / nex-n2-pro / step-3.7-flash #6 - 100k usd GB300 station / 8x 6000 / 4x H200 / Mix GLM-5.2 / Kimi-K2.7 ——— Let’s keep the Internet free thanks for reading

@RhysSullivan Lmao

last one

Crazy to wake up to this news. If this trend continues, access to frontier intelligence could become increasingly restricted. Running local models suddenly makes a lot more sense, even if they’re not quite at the level of the latest cloud models. This feels like a major shift. Does that mean Opus 4.8 and GPT-5.5 could end up being the most capable widely accessible models before stricter controls arrive? The next few weeks are going to be very interesting to watch.

@thdxr Password123!?

what's your password?

@durov So you mean I can control my Hermes agents via my Apple Watch?

🔄 Telegram’s new update is out. Lots of great new features. Most importantly: the announcement is actually fun to read. telegram.org/blog/watch-app…

@thsottiaux Tibo, Codex is very slow today. Are you launching 5.6 or what’s happening?

Heard your (amusing) feedback that it was at times annoying to receive a reset of your Codex usage without warning. Next time we press the button you will get to choose when it actually applies. Happy codexing.

@petergyang @_nilni Running it in fast mode today, but it's extremely slow. They might be cooking 5.6 right now 🧑‍🍳

@_nilni i have on fast mode all the time

These AI models remind me of RPGs tbh

@OpenAI Codex is extremely slow today. Are you planning on releasing gpt-5.6?

We heard you wanted to use Codex rate limit resets on your own time. Starting today, we’re rolling out the ability to save rate limit resets to use later. We’re starting Go, Plus, Pro, and Business users with one free reset:

@CastAsHuman It takes forever today. They might release gpt-5.6 really soon...

Is it just me, or is Codex a lot slower right now?

@benln When Stockholm ser?

Cafe Cursor is booked in 26 cities over the next few weeks: Phnom Penh, Manila, Oslo, Dubai, Frankfurt, Belgrade, Lusaka, Guangzhou, Tirana, Shenzhen, Sofia, Cebu City, Hangzhou, Baku, Bogota, Lima, Malaga, Beijing, Cumbaya, OKC, LA, Philly, Eldoret, Leon, Tallinn, Culiacan