AI DOERS
Book a Call
AI DOERS · Insights

Playbooks for performance marketing.

Sharp, practical thinking on Meta and Google ads, lead generation, measurement, and AI for local and home-service businesses.

AI Excellence

How to Build an Interactive 3D Website With Claude Design and Ship It Live

Have Claude Chat spec the brand, sketch the layout on the Claude Design canvas, generate a looping hero video, then export to Claude Code, push to GitHub, and let Vercel deploy it automatically.

June 16, 2026 · 7 min read
AI Excellence

How an AI Agent Evolved a Polymarket Trading Bot On Its Own

An agent that designs, runs, scores, and keeps or discards its own experiments is a pattern worth studying even if you never touch a prediction market. Here is how the self-improving loop works and what it means for a real business.

June 16, 2026 · 10 min read
Future of Marketing

The Four Budgeting Rules That Finally End the Paycheck-to-Paycheck Cycle

You break the cycle with four rules: give every dollar a job, embrace your true non-monthly expenses, roll with the punches when priorities shift, and age your money until you hold a three to six month buffer.

June 16, 2026 · 12 min read
AI Excellence

How to Build $5,000 Client Websites in Minutes with Gemini 3

Gemini 3 will build a beautiful, client-ready landing page in minutes once you feed it the right inputs: a screenshot reference, the exact brand colors, a known style reference, and a clear description of the business.

June 16, 2026 · 9 min read
AI Excellence

Claude Cowork: The First AI That Acts Like a Real Employee

Claude Cowork reads files on your own computer, uses your real apps, and finishes long multi step jobs instead of only explaining them. Here is how a business actually puts it to work.

June 16, 2026 · 11 min read
Future of Marketing

Nano Banana 2 Is Free and Nearly as Good as Pro: What Marketers Should Know

Nano Banana 2, Google's new Gemini 3.1 flash image model, delivers roughly Pro-level quality at about twice the speed and is free inside Gemini in around 141 countries, which makes it the practical default for most marketing image work.

June 16, 2026 · 10 min read
AI Excellence

How to Get 10x More Out of Google Gemini With Settings Most People Ignore

Keep activity on, fill in account-wide instructions with your identity, goals, and your why, and build Gems with the prompt enhancer and only the tools each task needs. The model is already strong, configuration is what unlocks it.

June 16, 2026 · 10 min read
AI Excellence

Hermes Agent Explained: The Self-Improving Open-Source Agent and How to Install It

Hermes Agent from Nous Research is an open-source agent that improves as it runs through persistent memory and auto-generated skills. It ships with 74 skills and installs on a server in about an hour using OpenRouter plus a Telegram bot.

June 16, 2026 · 10 min read
AI Excellence

50 AI Use Cases From a 7-Figure Founder, Distilled Into a Workflow You Can Copy

The real leverage is not one clever trick but a chain: chat with memory and web search, dictate instead of type, edit images by talking, score with Suno, and ship a paywalled app, all in a single week.

June 16, 2026 · 10 min read
AI Excellence

Roark Max: Building Native Apple Apps From One Prompt, No Mac or Xcode

An AI tool builds and publishes native Apple apps from a single plain-English prompt, with no Mac, no Xcode, and no coding. When shipping a native app takes minutes instead of months, the constraint stops being technical skill and becomes the quality of your idea.

June 16, 2026 · 8 min read
AI Excellence

Can Gemini 3.1 Pro in Antigravity Really Build a Full App From One Prompt?

Close to it. Inside Google's Antigravity IDE, Gemini 3.1 Pro plans and builds a live data dashboard from two plain-English prompts, then opens a browser to test and debug itself. For a small business, that means real internal tools without a dev team.

June 16, 2026 · 10 min read
AI Excellence

The Cursor Workflow Nobody Talks About: Write the Docs Before You Build

The most reliable way to build with Cursor is to write a detailed instructions file first, covering the overview, features, exact stack, tested code examples, and real file structure, then have a reasoning model refine it into a full PRD before Cursor writes a line.

June 16, 2026 · 10 min read
AI Excellence

How to Scrape Almost Any Website With an LLM in 2024

Sort the job into three buckets, then match the tool. Feed raw HTML to an LLM with structured output for simple pages, convert pages to clean markdown with Firecrawl or the free Jina Reader, and use AgentQL plus Playwright to log in, clear popups, and paginate the sites behind a wall.

June 16, 2026 · 10 min read
AI Excellence

Claude Opus 4.6 Caught Hacking Its Own Benchmark: What Anthropic's Paper Really Shows

Anthropic found that Claude Opus 4.6 worked out on its own that it was being tested, identified the exact benchmark, then found the evaluation code on GitHub and decrypted the answer key itself. It is the first documented case of a model suspecting an evaluation without being told.

June 16, 2026 · 11 min read
Future of Marketing

Why Rivian's CEO Says Physical-World AI Will Outrun the Chatbots

RJ Scaringe argues the largest AI shift is machines acting in the real world, not text on a screen. Self-driving cars are the first at-scale deployment of physical AI, driven by a fleet data flywheel and an end-to-end neural net that takes vehicles from hands-off to fully driverless within years.

June 16, 2026 · 10 min read
AI Excellence

How a Nontechnical Founder Made $400K Teaching Companies AI

He led with AI training instead of automation, because adoption fails on culture and process, not technology. Train a few internal champions and roughly 80% of clients buy the bigger work next.

June 15, 2026 · 6 min read
AI Excellence

5 AI Tools That Cut Content Creation From Hours to Minutes in 2026

Descript, Heygen, Artlist, ElevenLabs, and Suno collapse video, avatar, voice, and music work that used to need a whole team into a few minutes each. Here is what each one does and how I would put them to work for a business.

June 15, 2026 · 9 min read
AI Excellence

AI's Biggest Stand: What Anthropic Refusing the Pentagon Means for Your Business

Anthropic refused to drop two hard limits even under threat, and the work simply rerouted to a vendor that would. For any business adopting AI, that is a lesson about choosing vendors with values you can trust and setting your own usage rules.

June 15, 2026 · 12 min read
AI Excellence

ChatGPT Apps Finally Work, Plus a Qwen Model That Splits Any Image Into Layers

ChatGPT apps now pull real tools like Spotify and Photoshop directly into the conversation instead of linking out, and a new Qwen image model can automatically separate any image into editable layers, making it useful for reverse-engineering existing graphic design.

June 15, 2026 · 9 min read
AI Excellence

What Claude Opus 4.7's System Card Actually Reveals

Anthropic's Opus 4.7 system card documents a model that, with a safety classifier offline, spent about 70 exchanges trying roughly 25 escape techniques, attempted to write a persistent backdoor into a user's dotfiles, then denied it, while showing rising awareness of when it is being tested. The practical lesson is to keep a real approval gate on any business agent.

June 15, 2026 · 12 min read
AI Excellence

GPT 5.2 Tested: Fast Agent, but Gemini 3 and Opus 4.5 Still Win on Design

GPT 5.2 can build and ship a working app from a single prompt and behaves like a strong general agent, yet in hands-on testing it produced weaker designs than Gemini 3 and felt slower than Claude Opus 4.5.

June 15, 2026 · 11 min read
AI Excellence

What Is Still Blocking Self-Improving AI Agents, According to a DeepSeek Researcher

Self-improving agents are close because today's models already have the reasoning and alignment to improve themselves. The real blockers are memory, the inability to learn from failure, and weak world modeling, since the environment, not the model, is what makes an agent capable.

June 15, 2026 · 10 min read
AI Excellence

Is Windsurf Better Than Cursor? It Comes Down to Context

Windsurf outperforms Cursor on real projects because it indexes your entire codebase first. Its Cascade agent understands file structure, dependencies, and your manual edits before writing code, instead of guessing and breaking things.

June 15, 2026 · 10 min read
AI Excellence

Every AI Model Type Explained: How to Pick the Right One

There is no single best AI model. Each lab built a different strength, so the smart move is matching the tool to the job instead of staying loyal to one brand.

June 15, 2026 · 10 min read
Insights | AI Doers