AGI is Here. Anthropic Just Proved It.

In This Issue

• Anthropic published data showing Claude succeeds at open-ended tasks 76% of the time. The case for practical AGI is already closed.
• Fernando Gómez charged 3,200 euros plus 299 per month to automate WhatsApp lead qualification for a real estate agency in Málaga.
• Nigam Goyal landed his first AI client by selling ownership, not a subscription. The playbook is simple and it works.

New Video:

AGI is Here. Anthropic Just Proved It.

Anthropic just dropped a report called "When AI Builds Itself," and it's the strongest argument I've seen that AGI is already here. By the definition that actually matters. Not the sci-fi definition. The practical one: can you hand a model an open-ended problem with no clear answer and have it go figure it out on its own? Six months ago, Claude's open-ended task success rate was 26%. It's now 76%. The average task has gone from 4 minutes to 12 hours. Engineers working alongside it are 8x more productive. And more than 80% of the code Anthropic ships today is written by Claude.

The mistake most people make: Still using AI for one-off tasks. Ask a question, get an answer, move on. That was fine in 2023. The people pulling ahead right now are handing off entire workflows and only stepping in to apply judgment.

How this helps you: The scarce skill right now isn't prompting or coding. It's knowing what to delegate, what output to trust, and when to push back. That skill compounds. The sooner you build it, the wider your lead gets.

Winning Is All We Do

How Nigam landed his first AI client by selling ownership, not a subscription

What happened: Nigam Goyal just closed his first real AI client: a 20-person retail brand-tech company drowning in repetitive manual ops work. Instead of pitching a tool they'd depend on him to maintain, he sold them something they'd fully own. He built it on their infrastructure, with their API key. MCP server (Docker) + Claude Desktop on top, knowledge base living in their own git repo. Before sending the proposal, he built a working demo of one workflow. He shipped 2 complete production pilots end-to-end, then documented framework scaffolds for 7 more use cases, ready to activate later. Land, then expand.

The opportunity for you: The insight that closed the deal: sell ownership, not dependency. Bootstrapped buyers trust you more when you hand them the keys. The follow-on work comes anyway, on their terms. Charge a small paid discovery phase first to lock scope before you build. And before you pitch: build a 2-3 hour sandbox demo. Nigam says it wins the same deal as a 15-hour pitch.

Stories like this happen every week inside AIS+. AIS Live (July 11-12) is where you see exactly how: practitioners walking through real projects on stage, from first outreach to signed invoice.

In the News

The two AI labs powering your stack just filed for IPOs

The news: Both Anthropic and OpenAI have filed S-1 documents with the SEC. Anthropic filed on June 1 at a reported ~$965 billion valuation. The two companies whose models power most AI automation work today are both heading toward public markets.

How it affects automation: Public companies face shareholder pressure to grow revenue. For builders, that means API pricing, rate limits, enterprise tiers, and feature prioritization can all shift after an IPO. This isn't a reason to panic. It is a reason to know exactly what you depend on. Document your stack now, while switching costs are low, and have at least a mental fallback for each service you rely on.