Andrew (@AndrewGiorgiAI) profile picture

I build AI agent teams that automate businesses front & backend operations.

Joined March 2026
Photo by Andrew (@AndrewGiorgiAI)Photo by Andrew (@AndrewGiorgiAI)Photo by Andrew (@AndrewGiorgiAI)Photo by Andrew (@AndrewGiorgiAI)Photo by Andrew (@AndrewGiorgiAI)Photo by Andrew (@AndrewGiorgiAI)Photo by Andrew (@AndrewGiorgiAI)Photo by Andrew (@AndrewGiorgiAI)Photo by Andrew (@AndrewGiorgiAI)Photo by Andrew (@AndrewGiorgiAI)
I built AI agent teams that runs 1 of my businesses & 2 of my client's businesses. Scheduling. Onboarding. QA. Order tracking. All of it. What used to eat 60+ hours of my week now runs without me touching it. Here's exactly how I set it up 🧵
5
16
1,227
Absolutely huge.
Claude Code can read and answer your Hermes Telegram, Discord and Slack chats… through one MCP server Introducing the official mcp-serve plugin for Hermes that is now in the plugin catalog 🪽 it's pretty easy to use... Install the plugin and point your client at hermes mcp-serve check this out, left a few commands in the replies
23
Build in silence and let your results speak for you.
40
Lmao if you know then you know when it comes to what he is saying here. This is so fucking true.
there's a simple way to 100x your AI outputs and most people will never do it... agents like Claude Code can spawn sub-agents that run tasks at the same time most people use this for basic stuff the real play is having 3-5 sub-agents review Claude's work from completely different angles... using ANOTHER model (ChatGPT, Grok or anything really) but don't use fake "expert personas" you invented in a prompt, you need real frameworks from real people how to build these expert profiles: option 1: use someone so well-known their thinking is already in the training data > have a Karpathy-style agent tear your systems apart > build a profile based on Ogilvy's principles to roast your copywriting option 2: go deeper with Obsidian and Hermes / Grok bot > feed it an entire youtube channel from someone you trust > or their blog, their newsletter, their podcast transcripts > prompt it to extract an expert card with their frameworks, decision patterns, and principles now upload that material into your agent what happens next: Claude generates the first draft > sub-agent 1 reviews it through expert A's lens > sub-agent 2 stress-tests it with expert B's framework > sub-agent 3 catches what the others missed > all running at the same time Claude thinks through every piece of feedback and rebuilds you only see the final version... the one that survived 3-5 rounds of real scrutiny this is close from getting the absolute best output for a specific task
46
Andrew retweeted
anyone claiming GPT Astra is AGI is either trying to sell you something or clearly has zero understanding of AI
61
24
9
496
31,878
I don’t listen to anyone about AI unless it’s @EXM7777 or @jun_song . Don’t listen to these people who immediately jump on the bandwagon & say “Grok Bot is the best AI there is & I’ve ever used” when the damn thing hasn’t been out long. It’s ridiculous how these people operate. Listen to the people who ACTUALLY build with the tech, not make basic “8 use cases that will change your life” YouTube videos. Shit is ridiculous. Use cases you’ll never fucking use.
124
Series: Fable 5 / Hermes Agent / AI Agents / Memory / Local-First Operating Systems My Hermes Agent Couldn't Learn Anymore. Fable 5 Found Out Why Timestamps: 00:00 The memory was 99.99% full 00:47 Background 01:45 Fable 5 architect, Hermes operator 02:08 Why a full memory breaks the agent 03:22 Paid memory was 91% noise 04:00 A credential in cloud memory 04:33 The fix: 2 AIs, 11 rounds, 2 days 05:09 Auditing 111 memory entries 05:52 The 25 question benchmark 06:37 The boring stack won 07:21 Raw 20 of 25, adjusted 25 of 25 07:58 Retrieval beats reasoning 08:15 My 3 layer memory setup 09:06 Inside the Obsidian vault 09:46 The capture net 10:21 Audits, backups, and the defect we caught 10:56 The 6 step playbook 11:52 Cleaner, safer, searchable
4
1
1
433
No surprise at all. 😂
We've reset 5-hour and weekly rate limits for all users.
1
183
Andrew retweeted
We're extending access to Claude Fable 5 on all paid plans through July 12.
5,561
8,941
8,235
80,393
22,933,313
For operators, that's the real value. Fugu isn't "smarter answers." It's workflow governance: catching bad assumptions, comparing implementation paths, spotting missing gates, and stopping false-complete dashboards before they cost you a week.
1
73
Hermes already has the tools: files, memory, cron jobs, browser, terminal, Google Drive, Buffer, X, dashboards, and subagents. Fugu doesn't need to be the whole system. It's the expert review layer sitting inside it.
69
Fugu Ultra clicks inside Hermes once you stop treating it like a prompt box. Hermes is the operating environment. Fugu is the high-reasoning reviewer you bring in when the quality of the plan actually matters.
51
Reserve Fugu for the leverage points: new automations client-facing systems launch workflows security and safety reviews high-stakes strategy That's where pairing Hermes with a hard reviewer pays for itself.
50
The mistake: sending every tiny task to the most powerful model. That's slow and expensive. Keep normal execution on Hermes, GPT, or Codex. Bring Fugu in for audits, architecture calls, proof reviews, failure analysis, and "what are we missing" checks.
101
The loop I run: Hermes executes. Fugu reviews. Hermes fixes. Fugu verifies. One agent does the work, another stress-tests it, and nothing gets marked complete until the proof exists.
38
Most people waste Fable 5 by making it do cheap work. I Found the Best Way to Use Fable 5 With Hermes Agent Timestamps: 00:00 Why most people use Fable 5 wrong 01:09 Andrew's real AI agent context 01:47 The copy-paste mistake with Fable 5 02:23 Claude Code as the bridge to Hermes 03:00 Fable 5 does not replace Hermes 03:16 Hermes executes, Fable 5 reviews 04:20 The practical Fable 5 workflow 05:22 When to use visible review 06:04 How this compressed a week of work 06:52 What the setup looks like 07:27 Backend workflow vs operator visibility 08:23 How to set this up from scratch 09:28 Why the stack is a major unlock 09:58 Give each model the right job 10:41 Subscribe and join the community
1
134
The mistake is using Fable 5 for cheap execution work. The 07:45 client example shows why placement matters. Full breakdown: youtube.com/watch?v=1evLG-f1…
64
Most people waste Fable 5 by making it do cheap work. I Found the Best Way to Use Fable 5 With Hermes Agent Timestamps: 00:00 Why most people use Fable 5 wrong 01:09 Andrew's real AI agent context 01:47 The copy-paste mistake with Fable 5 02:23 Claude Code as the bridge to Hermes 03:00 Fable 5 does not replace Hermes 03:16 Hermes executes, Fable 5 reviews 04:20 The practical Fable 5 workflow 05:22 When to use visible review 06:04 How this compressed a week of work 06:52 What the setup looks like 07:27 Backend workflow vs operator visibility 08:23 How to set this up from scratch 09:28 Why the stack is a major unlock 09:58 Give each model the right job 10:41 Subscribe and join the community
117
Fugu Ultra isn't another prompt box. It's an orchestration system. A prompt goes in, then it routes, plans, hands work to a worker, verifies, and synthesizes the answer. It picks the right model for each step instead of forcing one model to do everything.
1
68