ianpurton

It's very easy to install and add models.

Because writing prompts is hard.

Most AI coding tools can generate a Todo app from a small prompt. This is because that problem is well understood.

When you try to use AI coding tools on your own projects you need to start writing a prompt that teaches the AI about your current architecture and decisions.

So the initial prompt is large.

Often the task needs knowledge of other files in your project. You can add them by hand or some AI tools will search the code base.

The prompt is now huge.

When you run that prompt you may or may not get what you expected.

So now the issue is how much time do you spend getting the prompt correct vs just writing the code yourself.

This area is brand new and there are very few resources on how to use AI coding tools effectively.

I have yet to see one demonstration of effective AI coding tool use on a project of reasonable complexity.

1. Fix it in your company.

2. Create a Youtube video of how you fixed it.

3. Spread that video far and wide.

4. If you get traction consider adding an offering

Be the go to person on Youtube for infrastructure cost savings.

The enterprise is a shitty place to find recognition. Seek it elsewhere.

I've been coding professionally for 30 years.

I'm probably in the same place as the author, using Chat-GPT to create functions etc, then cut and pasting that into VSCode.

I've started using cline which allows me to code using prompts inside VSCode.

i.e. Create a new page so that users can add tasks to a tasks table.

I'm getting mixed results, but it is very promising. I create a clinerules file which gets added to the system prompt so the AI is more aware of my architecture. I'm also looking at overiding the cline system prompt to both make it fit my architecture better and also to remove stuff I don't need.

I jokingly imagine in the future we won't get asked how long a new feature will take, rather, how many tokens will it take.

The developer tools market is REALLY difficult.

Nearly every architecture decision is "which open source tool can I use to solve [problem]".

No one wants to pay for anything and that's ok.

What is that?

Do some people have examples of interactivity they were able to replace with Htmx?

For me, I've been able to get turbo style links with the boost to get nice page transitions. I can also see how I could use the class-tools extension to enable buttons to open dialogs etc.

I'm curious to see when people say they needed an SPA for interactivity, what interactive features Htmx can already do and when do we need to break out some JS.

An example that I think needs JS is a copy paste button.

I'm using Dioxus on the server side and have written about it here. https://rust-on-nails.com/

The way you write components is really great and the overall developer experience is very good.

You didn't say if you'll have a front end.

Supabase works best with single page apps. However a lot of the frameworks are moving towards server side react, so the need for having a rest API is going away.

So I would actually say use nextjs and get an AI such as bolt to generate an MVP.

If you are running AI such as LLM's you need GPU's. LLM's are constrained by memory bandwidth not CPU and GPUS have a ton of memory bandwidth.

Postgres is always a sold choice and RLS is defence in depth anyway.

I can see a lot of benefits of having 1 API to access a lot of functionality.

For example instead of integrating with a message queue I can just do an INSERT this is great. It lowers the friction.

Vector search is a no brainer too. Why would I have 2 databases when 1 can do it all.

Using Postgres to generate HTML is questionable though. I haven't tried it but I can't image its a viable way to create user interfaces.

I find the "in 1 file" message confusing.

Does it mean sqlite, so just 1 file which is the database?

I want an AI to take issues and turn them into PR's.

I want the PR's to be so good and well tested that I can actually approve them and commit them to the project.

Lighter than the existing light weight kubernetes alternatives i.e. k3s :)

DHH in the past has said "This setup helped us dodge the complexity of Kubernetes"

But this looks like somehow a re-invention of what Kubernetes provides.

Kubernetes has come a long way in terms of ease of deployment on bare metal.

Bionic - https://bionic-gpt.com

Built around teams so you can build and share AI assistants based on your data.

Because PoW would only slow the attack.

Bots networks are often malware running on infected machines, so there's no cost to the hacker.

Captcha is supposed to stop the attack altogether.

> Products are built to fit with channels. Channels do not mold to products.

Some great advice in those articles. Thanks for sharing.

I've solved the same problem but used Kubernetes namespaces instead.

Each customer gets their own namespace and a namespace is locked down in terms of networking and I deploy Postgres in each namespace using the Postgres operator.

I've built an operator for my app, so deploying the app into a namespace is as simple as deploying the manifest.

> We wrote some code to automatically generate an HTTP API from the corresponding gRPC API

It's possible to do this without writing code. https://github.com/grpc-ecosystem/grpc-gateway

You can even get it to generate Open AI style documentation.

That's a difficult one. Here's what I would look at...

- Keyword research. What are people looking for.

- Content marketing i.e. blog posts, tutorials

- Google Ads

- A free widget, something to attract links.

- Cross posting interesting content to subreddits i.e. design

- Youtube - creating content or sponsoring content.

- Outreach - depends on your market but maybe linkedin.

- Buying high traffic websites and integrating the content into your own site.

https://bionic-gpt.com Just signup and start adding data.

> I need a team of freelancers to build the MVP.

Why a team? How about just one.

> where do you find effective and affordable freelancers?

upwork.com

Ideally you'll have good requirements for what you want. If it's a web product create a wireframe. If you can create a clickable HTML wireframe even better.

If there's an open source equivalent just use that.

It's fine to be in AI.

My takeaway from the article is instead of being a Gen AI startup be a Gen AI startup for a specific use case.

Munich has a great surf and skate culture with a new wave opening soon. https://surftown.de/en

There a lot of trails if gravel biking is your thing and plenty of sailing opportunities on the big lakes.

So like any city its more about what you make of it.

> All government ministers' pay, including that of the mayors, should be tied to that of nurses for the next 100 years.

A brilliant policy.

The economics of purchasing new are fairly simple, you walk out of the store with something worth more than you paid for it.

The people in the secondary market take all the risk by paying an inflated price that may go down and the risk of fakes.

It reminds me of the Rolex market.

The article makes sense.

In a way experience is knowing what things will cause friction and having solutions to overcome them. Friction reduction.

A recruiter gets paid a commission if they put you forward and you get the job.

So they only put candidates forward if they have a good chance of getting the role.

You can use them to get feedback on the current market and how they think you might fit into it.

It's worth a try.

Take a look at Bionic https://bionic-gpt.com/

You can check out https://github.com/bionic-gpt/bionic-gpt

Basically I put db in it's own crate then crates for controller and another for pages.

The folders for each section of the web application.