Anthropic's internal loop engineering playbook just got leaked.
And it's the most valuable AI guide I've read all year.
This guide is packed with info, and there are five things you need to know if you want to maximize your AI productivity with loops (save this):
1. You should structure every loop around these 5 principles:
• Discovery → Let the agent find its own work (CI fails, issues, commits)
• Handoff → Give every task its own isolated git worktree
• Verification → Never let the generator grade its own work
• Persistence → Always write state to disk (markdown or board)
• Scheduling → Run it on a timer so it works while you sleep
2. Separate Generator from Evaluator (most important rule)
Use two agents: one writes, the other is a skeptical judge who assumes the code is broken.
Make the evaluator act (run tests, click buttons, take screenshots) - this is what actually stops bad output.
3. Build with these 6 parts:
• Automations (the timer)
• Worktrees (safe parallelism)
• Skills (permanent project knowledge)
• Connectors (talk to GitHub, Linear, etc.)
• Sub-agents (generator + evaluator)
• Memory (state files that survive between runs)
4. Things you must watch out for with loop engineering:
• Verification debt (use verification agents)
• Losing understanding of your own codebase (start fresh if needed)
• Token costs exploding (solution below)
• Cognitive surrender (don't stop thinking because “the loop handles it”)
5. Solving token costs
Loop engineering can be crazy expensive.
I recommend you use an 80/20 "barbell" approach to loop engineering.
For your most complex tasks that require the best intelligence, use expensive models (Opus).
For the remaining 80% of your tasks (the gruntwork), use cheap, open-source models within the Claude Code harness (GLM-5.2 is great for code execution).
Save these 5 rules so you don't forget them.
The robotics trade could be the most asymmetric bet of the next decade.
Robotics company investments just smashed a new all-time high (~$16B and climbing).
These are the top robotics companies/tickers I'm watching for a 10x (nfa):
In crypto, every cycle has heroes that power the bull market cycle.
These are no ordinary heroes!
These are a special type of hero!
THESE ARE LEVERAGED HEROES.
Only thing is that the market hates leverage and so these same leverage heroes are destroyed and eliminated in the brutal bear market cleansing!
The bigger they were in the bull market, the more leverage they created , the more brutal the collapse in the cleansing.
And only when the leverage is totally gone , the new cycle begins....
"Leverage" is not only borrowing from exchanges. It is he creation of new funds without a 1:1 backing!
It's always the same:
1.Assets appeared larger than they really were.
2.Confidence attracted more capital.
3.Rising prices validated the structure.
4.Falling prices exposed the mismatch.
5.Collapse destroyed purchasing power.
Last cycle,
- Do Kwon created a $40bn leverage from printed money.
- SBF leverage his client funds into a private VC fund.
- Alex Mashinsky leveraged his client funds into more leveraged client funds.
- Kyle and Zhu Su leveraged their investor funds to the hilt!
This Cycle's leveraged heroes...
Saylor? - He leveraged his MSTR stash to promise 11.5% a non- yielding asset.
Tom Lee?- He is now starting to leverage 5% of the ETH supply (more responsibly) with a pref.
Just look for the leverage and when totally gone and there is NOTHING left to squeeze, thats when the bear market will be over and the new leveraging cycle will begin.
Anthropic engineer:
"At Anthropic, 90% of our engineers use loops and ‘dreaming’ to build self-improving agentic systems.
сlose the agent loop. give an agent a way to verify its own output."
in a 30-minute session, an Anthropic team member explains how to build an agent that improves itself.
Claude + loops + dreaming + CLAUDE.md - that’s the secret.
Watch the talk, then save the playbook below.
THE SIMPLEST STRATEGY BEAT 98% OF TRADERS OVER 10 YEARS
No indicators. No complex systems. Just one rule.
2018, 2022, 2026 - same signal every cycle:
BTC below MA200 → buy and hold
BTC above MA200 past day 1,000 → sell
That's it. Everything else is noise.
I called the exact bottom at $16K three years ago and the $126K top in October both publicly, both before they happened
The next call is already forming
Turn on notifications - you'll understand why you needed to follow sooner
Anthropic engineers just showed how they build a full app from scratch, using a loop of agents
40 minutes from the team behind Claude Code
they used three agents: one to plan, one to build, one to judge, cycling until the app actually works
the winners won't have the smartest model, they'll have the best loop
watch it, then read the full guide on how to actually use loops below
I connected my entire business to Hermes x Obsidian.
Client notes, SOPs, meeting logs, business decisions - all of it.
My Hermes agent now runs automations I didn't even know I needed, and my agent self-evolves over time.
Here's exactly how I set it up:
Step 1. Create your business vault in Obsidian
Head to obsidian.md and download the app.
Create a new vault called "Business Brain."
Then dump everything in:
- Client notes and CRM data
- SOPs and processes
- Meeting notes and action items
- Business goals and OKRs
- Financial data and decisions
- Any context you want your agent to know
The more you feed it, the more powerful it gets.
Step 2. Connect Hermes to the vault
Open the Hermes desktop app and paste this prompt:
"I want to connect you to my business Obsidian vault, and have you act as my business orchestrator - do everything necessary to set up that connection now."
Hermes now has direct access to your entire business context.
Step 3. Let it build
Every time you add a new note, Hermes automatically ingests it.
Over time, this builds a living memory of your entire business context.
Start prompting your agent with things like:
"Based on everything you know about my business, what automations should we build first?"
You can even send prompts like:
"Every single night while I'm sleeping, ingest data from my business vault and autonomously build automations."
This is one of the most powerful *yet simple* automations you can run.
Bookmark this so you don't forget it.
R.I.P. rebuilding your GTM stack from scratch every session.
A complete Claude Skill Library can replace a $15,000/month agency retainer.
It is not as easy as hiring someone else to do it.
But if you start today, you can have 130+ skills loaded into Claude Code covering every GTM job by end of this week.
I usually charge $299 for access to this library but today, it's free.
Like this post + comment 'Agents' and I'll DM you the entire skill library for free.
(Must be following, or I can't message.)
Taking this down in 48 hours.
This model is insane at design.
I asked GLM 5.2 (left) and Opus 4.8 (right) to build me a landing page and you can't even tell the difference.
GLM cost $0.06 while opus cost $0.49. More than 6x cheaper while being faster + more token efficient.
Another win for open source AI.
I was wrong
I've been saying for months that open source AI models are 6 months behind frontier
They caught up. GLM 5.2 is as good as Opus 4.8
This changes everything. If you run GLM 5.2 locally no government can take it away. You become sovereign
And even if you run through APIs, its a fraction of the cost
The battlefield is different now. If open source is as good as frontier, and people have cheaper alternatives, governments can't be as quick to regulate. It will destroy the frontier AI labs
All of this is such a massive win for the people
If you are not paying attention to local models yet, you are making a tremendous mistake
Developers can now build on Yoco with the new Yoco Developer Hub 👨💻
Yoco has opened its infrastructure to developers, giving them access to APIs, documentation, and an MCP server to build custom applications, integrations, and workflows on top of the platform.
This helps developers:
• Build commerce and payment solutions without starting from scratch.
• Use AI tools to create integrations and applications faster.
• Access the same infrastructure powering over 200,000 South African businesses. #YocoNext
Hermes just partnered with Stripe.
Your AI agent can now buy products, pay for API calls, and manage your finances through Stripe skills.
Here's exactly how to set it up:
1. Install the specific Stripe skills you want (they're optional and modular):
https://t.co/FnhGC51rsR
2. Connect your Stripe account
Run the skill or ask Hermes directly
(e.g., "Connect to Stripe" or "Set up payments").
It will guide you through OAuth/API key auth securely (via Stripe CLI, Composio, or built-in flows). No raw card details are ever exposed to the agent.
3. Set safety limits
Configure USD spending caps, approval thresholds, allowlists, and audit logs.
Example: Limit to $[x} per transaction or per day; require approval for anything larger.
Hermes has built-in safeguards (human-in-the-loop, temporary keys that auto-expire, etc.).
4. Test and use
Tell Hermes something like:
"Renew my domain and pay under $50"
"Book a flight to Lisbon next Thursday, aisle seat, under $400"
"Provision a Neon DB and pay for it"
"Pay this API call"
It handles checkouts, upsells, etc., within your limits. All activity shows in your normal Stripe dashboard with standard protections.
TIME TO BECOME BULLISH bitcoin:native !!
We are repeating 2022 bottom quite exact
✅ 200W hit
⏳then chop
⏳then Valhalla
By Q4 we should reach the end of chop
2027 valhalla
Since October 2025 I am bearish
And to be frank I still expect a bit more downside
BUT THE BEAR WILL BE COOKED SOON
VERY VERY COOKED
One thing the Fable ban has made clear is that open source/decentralised AI is essential.
No singular party should have full control.
This could serve as massive accelerator of the trend.
Fable was only available for 72 hours.
In that time, I was able to:
• Create a workflow for editing my YouTube videos
• Build a fully autonomous clipping system
• Overhaul my entire BD pipeline - better data enrichment/accuracy
Next I was going to start on my executive assistant build (compressing every single manual workflow into loops).
It was truly insane.
I can't wait to get access again.
This just opened Pandora's box. There is no going back.
This is one of the craziest headlines you'll read all year.
The US government has suspended the use of Claude Fable 5 & Mythos 5.
My thoughts on the ramifications:
- There is a strong motive to overturn the ban, as the ramifications for equities are significant. Trillions of dollars of data centers/contracts rely on the AI trade - free markets don't like intervention.
- Raises concerns over favoritism, clearly favors competitors - this likely buys OpenAI/Google more time.
- Sets a worrying precedent in terms of control (regulation is important, but I can't help but feel the best models will be controlled by governments and big corporations - they're afraid of the little guy having too much power).
- Open-source models are going to accelerate and drive cost down (there is growing demand) - meanwhile, frontier models are only getting more costly. Has fascinating knock-on effects for the AI trade.
- A reminder that if you don't own your own models/compute, you're trading sovereignty for convenience. Local models will become a massive trend.
I use Claude every day to run 3 companies and a 30-person team.
It pains me to watch people use the most transformative technology ever built like it's just a fancier Google Search.
Claude can change your life, but you have to use it correctly.
How to master Claude (from zero)
This is a historic day for AI.
Claude Fable (Mythos) was just released, and it's insane.
I can't believe Anthropic made this public.
This video covers my first impressions, real use cases, and full breakdown of the world's most powerful AI ever: