Domain Expertise + AI is the new baseline. Learn how to combine your expertise with AI to build prototypes, tools, and workflows without becoming a developer.
Most of us were taught to plan everything up front, then build exactly what the plan says.
When you're building products with AI, that's backwards.
Some ideas you simply can't judge on paper. You have to build them to find out. So the plan you locked down on day one quietly becomes the thing holding you back.
I wrote up a different way to work.
I call it Prototype Driven Development (PDD).
You keep your plan, but hold it loosely. When you hit a decision you can't make on paper, you spin off a fast, throwaway prototype to answer that one question, then fold what you learn back into the plan. The code is disposable. The learning is permanent.
It's how I build my own apps.
Read the full write-up here:
https://t.co/oGnsBg6eA8
And if you want to see it live, I'm running a PDD session at the I Build with AI Meetup on Friday, June 26 in Fullerton, CA.
First the talk, then a hands-on workshop where we build a small app together with Cody Product Builder using PDD as the backbone.
Bring your laptop, an AI agent, and an idea you've been wanting to build.
RSVP here:
https://t.co/gARqir0ckU
What's the decision you've been stuck on that you could settle this week by building a small prototype and just finding out?
@AnthropicAI@ClaudeDevs@claudeai Why can you set the advisor to Sonnet in Claude Code, but the messages API explicitly says you can only use Opus 4.7 and Opus 4.8 as advisors? But we set up sonnet 4.5 using sonnet 4.5 as executor and it worked. Also, in Claude Code, you can't have Haiku 4.5 as an executor for the advisor to run, but in the messages API, you can. Can you explain the inconsistencies?
Tonight is our first I Build with AI Buildathon in Fullerton!
Bring your laptop, your favorite AI tool, and an idea you want to build.
We’ll have starter prompts, helpers, pizza, prizes, and a room with people building with AI.
The goal is not to build something perfect.
The goal is to start building.
RSVP here:
https://t.co/ybBxqaCleX
#AI #BuildWithAI #VibeCoding #ProductBuilding #Meetup
AI gives me product-building superpowers.
This afternoon, I’ve been jumping between several projects:
- My new iPhone and Android game, soon to be released, built with Flutter
- Cody Product Builder, using markdown and Python scripts for agentic workflows
- Checkinto, an app for people to check into my meetup, added polls for voting, built with SvelteKit and Tailwind
All running in three separate Claude Code sessions (inside VS Code)
At the same time, I’m updating a buildathon document for my meetup tomorrow in ChatGPT.
And in another ChatGPT project, I’m planning my September trip to Japan.
All of this is happening literally at the same time.
I’m not saying it’s effortless. You still need taste, judgment, direction, and a clear sense of what you’re trying to build.
But wow!
Being able to move across this much variety in one afternoon (successfully) with AI helping is a pretty incredible feeling.
This is what building with AI feels like when it starts to click.
I finally launched my site for my Cody Skills!!!
https://t.co/Pu0MkBNfjv
Cody skills are AI agent skills that help turn ideas into shipped products. Built for knowledge workers, domain experts, and creators.
Get the full documentation for every skill. Plus download from there directly to use in your projects.
Check it out. Drop your thoughts below. Reach out to me directly via DM or [email protected] with comments or questions.
🚀 Cody Product Builder is now officially an Agent Skill.
When I first built Cody Product Builder, the Agent Skill standard did not exist yet. So I built my own activation layer to make it work across AI coding tools.
That standard exists now. With v2.0.0, Cody Product Builder is a proper Agent Skill.
What that means for you:
✅ One self-contained skill. No custom setup, no per-tool workarounds.
✅ Install once (in your home folder), use it across Claude Code, Cursor, GitHub Copilot, Codex, and more.
✅ On https://t.co/JrvQN0RSq1 (desktop or web), just add the .skill file.
What has not changed is the part that matters: the workflow. Cody still takes knowledge workers and domain experts from a raw idea to a finished, shippable product through a structured two-phase process, Plan then Build. Discovery, PRDs, plans, feature backlogs, versioned builds, prototypes, release notes. The full trail.
The goal has been the same since day one: let people who are not developers build real products with AI, with structure instead of chaos.
v2.0.0 just makes Cody easier to install and easier to take with you everywhere.
Cody covers the four things you actually do when building a product:
✅ Full planning to shape the idea
✅ Full build cycles to ship it version by version
✅ Prototype creation to test a theory before you commit
✅ Idea tracking to catch the good thoughts that show up along the way
And it supports both greenfield (new) and brownfield (existing) projects.
Free to use. Try it:
https://t.co/EPzSfh7rWb
#AgentSkills #AI #ClaudeCode #BuildWithAI #iBuildWithAI
Cody Product Builder v1.11 released!
You can now prototype your ideas.
Sometimes you have an idea but you are not ready to commit it to a plan or a build. You just want to see it, click around, and find out if it actually holds up.
That is what the new :cody prototype command is for.
Run it anytime, build a quick throwaway prototype, note what you learned and what you liked or disliked, then keep it or toss it. It is completely independent, not a phase, and it never locks you into anything. When you later plan or build, Cody quietly reminds you which prototypes you have so you can pull what you learned into the real work.
With this release, Cody now covers the four things you actually do when building a product:
- Full planning to shape the idea
- Full build cycles to ship it version by version
- Prototype creation to test a theory before you commit
- Idea tracking to catch the good thoughts that show up along the way
It is free, it is markdown-based, and it works with whatever AI coding environment you already use. If you have been wanting a calmer, more structured way to build with AI, take a look.
https://t.co/EPzSfh7rWb
When speaking about AI, we always talk about how fast it is at coding, at shipping, at pretty much everything.
But I noticed something else, and now that I noticed it, I see it constantly in human collaboration.
Friction.
I don't mean dramatic conflicts. I mean the small, ambient stuff that shapes how collaboration actually feels.
The tone shift after the third scope change. The way "I don't think that can be done" closes a conversation before it opens.
The pause before "we just don't have the time to do that," where you can feel the door shutting on an idea you haven't even finished explaining.
Ego. Mood. Defensiveness. Insecurity. Territorial thinking.
The reflex to dismiss a possibility before spending ten minutes exploring whether it's real.
These aren't character flaws. They're human.
Every one of us carries some combination of them into collaboration, including me.
And once you've worked without certain frictions, you start seeing them everywhere you used to accept them.
AI lets you work differently. And in working differently, it holds up a mirror to every collaboration habit you've built over your career.
Once you've worked that way, you can't unsee it.
👉 https://t.co/Jj80oUZNnl
We’re doing our very first "I Build with AI" Buildathon: From Idea to Demo on May 28 at 6:30pm in Fullerton, Southern California (exact location is in meetup page).
Think of it like a hackathon, but designed for builders, not just developers.
This event is for business professionals, knowledge workers, domain experts, product thinkers, curious creators, and anyone who wants to use AI to turn an idea into an app, prototype, workflow, mockup, tool, or something demoable.
We’ll provide starter ideas, example prompts, and practical guidance to help you get moving quickly. We’ll also have people available during build time to help answer questions, get you unstuck, and help you focus your idea into something you can demo by the end of the night.
Bring your laptop, an idea, and your favorite AI tool, such as ChatGPT, Claude, GitHub Copilot, Codex, Gemini, Cursor, or whatever you like using.
We’ll have pizza, possible prizes, and a chance for attendees to share short demos of what they built.
The goal is not to build something perfect.
The goal is to start building.
More details and free registration here:
https://t.co/ybBxqaCleX
#AI #BuildWithAI #VibeCoding #ProductBuilding #NoCode #AIProductivity #Meetup #BuildingWithAI #IBuildWithAI #AIEngineer #AIBuilder
I'm going to run a buildathon in my meetup, I Build with AI, at the end of this month (5/28/26). I'm looking for any vendors that would be interested in donating some prizes (AI tools, a 3 month access to your service, etc)?
Please reach out to me directly at [email protected] if you are interested!
You can check out the meetup here:
https://t.co/e7daUEHF1V
Some possible vendors that would be great if you can help out...
@OpenAI , @AnthropicAI , @ClaudeDevs , @github , @ollama , @figma , @Lovable , @vercel , @Replit , @cursor_ai , ????
If you can think of any other companies, please repost on your end with their name?
Thanks all!
"everyday we're trying to obtain more compute to pass on to you, we're sorry if it takes sometime but we're going to acquire as much as we can"
you heard the man
Agents can now create Cloudflare accounts, buy domains, and deploy
They can create a Cloudflare account, start a paid subscription, register a domain, and get back an API token to deploy code right away.
https://t.co/nPAbZs8jAM
I created this image with the newly just released @OpenAI ChatGPT Images 2.0. It's incredibly high fidelity. The text is amazing. I gave it my meetup event and my website as guidance for look and feel and it came up with this. I had 3 iterations with it.
Simply amazing.
While you are at it, join us in our upcoming (April 30th) meetup!
https://t.co/l9R9xU52bG
I wanted to understand /fast mode in Anthropic Claude Code. How does it work? Is it worth the extra cost? So I did a bit of research and wrote this short blog about it.
https://t.co/mCBga47zNj
#claudecode#anthropic#aiengineering#buildingwithai#agenticbuilder #agenticengineer
I was confused about the difference between 'effort' and 'thinking' in Claude. They sound like they do the same thing right? But after doing some deep research this weekend I found out they are not.
-> Effort is how many steps Claude takes from your question to an answer.
-> Thinking is how hard it reasons at each step, and whether it stops to reason at all.
Prior to Opus 4.7, these were separate dials. With the introduction of Opus 4.7, they have been fused.
My learnings can be found in this blog article:
https://t.co/n8iHEEJbmw
Hope it helps you. Tell me what you think and if you have noticed any differences with Opus 4.7 with regards to effort and thinking modes.
#claudeai #anthropic #claude #buildingwithai #productbuilders #ainative
I just released v1.6.0 of the @AnthropicAI@ClaudeDevs@claudeai Advisor Tool Playground and it's now officially MIT open source.
Quick refresher for anyone new to it: the playground is a hands-on sandbox for Anthropic's Claude Advisor Tool, the tool where a fast, cheap executor model (Sonnet or Haiku) can consult a smarter advisor model (Opus) mid-inference, only when it actually needs to. The idea: near-Opus reasoning at near-Sonnet cost, if you wire it right. Hard to reason about from code alone, so I built this to make every step visible: tokens, costs, timing, and the raw content each step produced.
What's in v1.6.0:
• Claude Opus 4.7 as advisor and executor (Anthropic now documents 4.7 as the only supported advisor)
• New xHigh effort level for Opus 4.7 only, recommended for long-horizon coding and agentic tasks
• Evaluator judge upgraded to Opus 4.7 (from Opus 4.6)
• Advisor caching dropdown now locks after the first message, per Anthropic's new "keep consistent mid-conversation" guidance
• Fixed a stale Opus 4.6 price that had been inflating in-app cost estimates
→ Try it live (bring your own Anthropic API key, no install):
https://t.co/AVgn0WBJvH
→ Fork, run locally, or contribute:
https://t.co/kKmUGyj2sx
If you're deciding whether the advisor tool makes sense for your workload, this should get you to an answer faster than the docs alone.
#AnthropicAPI #ClaudeAPI #OpenSource #LLM #AIEngineering
@madsf88 I hope it's useful!
https://t.co/Epvz6pIbmx
Learn the foundational building blocks of AI-powered apps through a block-based interface. Designed for people who want to build with AI without learning how to code.