mbm

SEEKING WORK - AI integration/implementation - Austin, TX & Worldwide Remote

Hi all, I'm a freelance AI consultant with 15 years of software engineering experience. I work with large companies, startups, and everything in between.

If you're beginning or considering an AI implementation in 2026, I'd love to chat. My specialities include: AI integration/implementation, complex backend development, APIs, AI strategy development, and legacy SaaS replacement.

Drop me a line. Happy to jump on a video call anytime to learn more about your needs.

Text: (512) 417-7608

Email: matt@goodsoftware.dev

LinkedIn: https://www.linkedin.com/in/matthewbmay/

Github: https://github.com/mzxrai

Ratatui is awesome! Just built a little chat client with it, tons of fun.

https://terma.mattmay.dev/

Memva, Claude Code for sales: https://memva.com/

Neat tool, thanks for sharing your work!

Ah true. So the push becomes even stronger for people to upgrade to the max plans, when several of the "apps" they're using are consuming their allocated Claude tokens. Brilliant.

Precisely. You click on a claude link, and suddenly it's, "You are now financially responsible for your actions from here on..." I'm sure they've spent a lot of time thinking through the ui/ux of this.

Agreed, it's an interesting model. I wonder what the approval ui looks like for the app end-user? Is it super clear to them that they're financially responsible for their usage?

Awesome anecdote, thanks for sharing this. How do product marketing and support engineering folks feel about the change?

Nice job! A couple questions for you guys:

- Did you personally end up getting offers using your own tool?

- Anything surprising you learned building it?

Interesting, so instruction-following troubles and also uncertainty by the model on which MCPs it should be using for a given task.

Interesting, thanks for sharing.

Awesome. Which servers do you use?

Thanks for the feedback! I just need some way then to indicate what each summary is without clutter. I've tried badges, text, and icons, but haven't had a lightbulb moment yet of a better way than just putting tl;dr. I agree it's not optimal!

Thank ya!

Added!

Indeed, it’s been a fun little project! Just pushed out a new version with 2.5 Flash and tl;dr’s.

Open sourced!

https://github.com/mzxrai/tinysums/blob/main/app/lib/ai/hn_s...

https://github.com/mzxrai/tinysums/blob/main/app/lib/ai/hn_t...

Hey mdp, open-sourced it this morning: https://github.com/mzxrai/tinysums

This was a fun little project, but unfortunately can't spend too much time on it. It's also quite expensive to run -- around $15 per day in Gemini summaries (I'm sure 2.5 Flash would be much cheaper, but unsure about the quality)!

Sure thing.

For the story summary: send a prompt to the Perplexity API (which can access URLs) requesting an extraction of the article's content (a sort of very detailed technical brief). Then, use Gemini to classify the results from the Perplexity API as valid or invalid (sometimes, Perplexity isn't able to access the URL but it doesn't reliably return the string I've requested it to return when it isn't). After that, send a more detailed prompt to Gemini requesting the story summary be generated with a specific style and format (Markdown).

For the comments summary: use the HN API to pull all comments, then prepare a Markdown document with a selected subset of comments (based on user karma, replies, and thread depth) if exceeding a given number of comments, otherwise all comments. Annotate comments with an index (eg, 1.1) to indicate the nesting to the LLM. Along with some formatting and stylistic guidance, send that to the LLM requesting summarization.

While the story summaries are generally static once generated, regenerate the comment summaries on an interval (this could be optimized).

Hopefully that helps a little!

Agreed. Great suggestion!

Paranoia does seem a tad strong, doesn't it?

Tech stack:

- Rails 8

- Embedded React frontend w/ Tailwind

- Sidekiq for cron jobs

- Perplexity API for story info extraction via URL

- Gemini 2.5 Pro for classifying story extraction results as valid/invalid

- Gemini 2.5 Pro for summarizing stories and HN comment threads (tried Claude but preferred the personality of Gemini surprisingly)

- Cursor with Gemini 2.5 Pro + Claude 3.7 Sonnet for codegen

Hey! Thanks for asking. In this case, I used:

- Cursor for IDE (Claude 3.7 Sonnet + Gemini 2.5 Pro for code gen)

- Rails app with an embedded React frontend (Rails for API routes / React for everything frontend)

- Deploy the Rails app to GCP Cloud Run with min_instances=0; use the tiniest GCP Cloud SQL instance for Postgres; and spin up a tiny cheap VM for running background jobs (same Rails app, just run in sidekiq mode instead of Rails mode)

DigitalOcean droplets (cheap servers) are a great option for doing it all on a single VM.

Thanks! It’s kind of long and rambling, but I’d be happy to share if you really want it. It needs some refining for sure.

Thanks! Anything else you’d like to see?

fun little side project using gemini 2.5 pro.

summarizes the top 30 HN stories and their comment threads. kinda nice when there's a huge thread to read the tl;dr.

cron job updates everything once an hour right now.

feel free to pick it apart!

Have been writing Rust servers with Cursor. Very enjoyable.

Beautiful article, thanks for sharing. Miss his writing.

Thanks for sharing! Do you do primarily internal dev?

Yeah, you really have to fast-forward 5 to 10 years. The first cars or airplanes didn't run particularly well either. Soon enough, we won't be able to tell.