Anthropic just dropped Claude Fable-5, a Mythos-class model that sits above Opus and is available to everyone right now, plus Claude Mythos Five for Project Glasswing partners only. I break down what the two models actually are, the pricing, why Fable-5 is only free on your plan until June 22nd, and the benchmark jumps in coding, knowledge work, and vision. I also revisit my earlier Mythos prediction and what I'm most excited to put Fable-5 to work on.

Subagents are one of the most powerful features in Claude Code, and most people barely scratch the surface. In this video I break down exactly what a subagent is, how it keeps your main context clean, when you actually need one versus when you don't, and how to build your own custom agents that get invoked automatically. By the end you'll know how to delegate work to a team of cheap specialist agents while one smart model runs the show, saving you money and getting better results.

Now I can post😁 I am currently at the day 4 of the #AISChallenge, looking forward to start the AI OS course!! I'm going to deploy my first automation, feel free to share any general advice :)

Here's the full breakdown of what it does: Runs daily on a schedule trigger. Pulls all leads from Google Sheets, filters out anyone already contacted, bad domains, wrong countries, SEO-polluted keyword entries. Passes each valid lead through a keyword cleaner that strips generic legal noise and extracts the actual practice area. Groq (Llama 3.3) writes a personalized cold email based on the specific practice area — personal injury, family law, immigration, criminal defense, whatever the lead actually does. The pain point is derived from their niche, not a fixed template. Email sent via Gmail. Thread ID saved back to the sheet. Second workflow runs at a different hour. Checks days since initial send. At day 4 — if no reply detected via Gmail Thread Get — it generates and sends follow-up 1 in the same thread. At day 9 — follow-up 2 (breakup email) with a direct Calendly link. Each step updates the sheet: Follow up 1 Sent, Follow up 2 Sent, error states handled separately. The reply detection works by checking messages.length > 1 on the thread — if someone replied, the system skips them. Tools: n8n · Groq · Gmail · Google Sheets

I've spent over 500 hours inside Claude's ecosystem, so I ranked every feature from D tier all the way up to S tier based on how much each one actually changes my day-to-day knowledge work and automation. Then I walk through my top 12 in order and explain exactly why each one earned its spot. Keep in mind I'm doing a lot of automation and knowledge work, not heavy software engineering, so you'll probably disagree with some of my placements. That's the point, you value these features differently based on how you use Claude Code.