AI Builder/Product Manager/Creator of Abu
I'm a product manager and an AI builder. I build indie products, write courses, and have delivered enterprise projects.
01About
I've spent more than $10,000 and over 1,200 hours coding with AI. My best-known work is Abu, an open-source desktop agent.
These days I deliver agents for enterprises, build the small tools I want for myself, and study ontology and the FDE delivery model.
02Products
I picked two to show here.
An AI desktop assistant that runs on your own computer. Tell Abu what you need and it does the work: reads files, runs commands, writes documents, builds reports. Free and open source, 120+ releases so far, with a commercial enterprise edition.
Learn more: AbuMake your cursor look the way you like. Turn any character into a full set of animated cursors, with click effects and text-selection highlights, on Mac and Windows. The cursor on this site is drawn after it.
Learn more: FunnyCursorI've also built
In the age of AI, everyone is a Builder. Just build it!
03For companies · FDE Delivery
I lead an FDE (forward deployed engineering) delivery team and a product development team. The delivery team has three roles, each with its own part in the five stages.
Goes on site, digs out the business problems nobody has spelled out, and defines the value this delivery should bring.
Writes production-grade code and turns the problem into a system that runs in the client's environment.
Builds the capabilities that keep coming up across projects into a platform, so each delivery gets faster.
Five stages in all. The team stays responsible until the client is actually using the system.
| Industry insight | Requirements discovery | Solution design | Build and deploy | Handover and operations | |
|---|---|---|---|---|---|
| Deployment Strategist | Involved | Involved | Involved | Not involved | Involved |
| Forward Deployed Engineer | Not involved | Involved | Involved | Involved | Involved |
| Software Engineer | Not involved | Not involved | Not involved | Involved | Not involved |
First the team works out how the client's industry runs, where the money goes, and how far its peers have got.
Interviews and research on site turn spoken requests into structured problems, and separate the needs people state from the ones they leave unsaid.
We decide where AI fits, where the data comes from, and how the result will be accepted. A first MVP is ready in one to two weeks to test the direction.
We connect the data, build the knowledge base and agents, deploy them into the client's own environment, and hook them up to existing systems.
After launch we watch whether the client really uses the system and whether it produces measurable business results, then keep adjusting.
04For individuals · Courses
I've written two illustrated courses and I'm preparing an in-person bootcamp. All of them are in Chinese.
For first-time builders. No coding background needed.
Six stages from idea to real users. Every chapter has a method and a template you can hand straight to AI, with six months of records from one real product running through the book.
For product managers and engineers who design or build agent products.
A product part and a technical part, 15 chapters each, with two design workshops: a research assistant and an enterprise customer-service agent.
For engineers and business people who deliver AI to enterprises.
One case runs through it: delivering a product-image generation agent to an e-commerce company, from research all the way to the final demo.
05Content
I share agent product development and AI productivity tips on these platforms, and document how I build my own products. Most posts are in Chinese.
06Partners
Media and content partnerships
07Contact
Enterprise AI agent delivery, content partnerships, product collaboration, courses and talks
Scan the code and tell me in a line what it's about. I'll reply as soon as I can.
No WeChat? Email meshawnsheno@gmail.com
On a phone, tap "Open" under a QR code. Inside WeChat you can also long-press a code to scan it.
