Spencer Jones
sjones6.bsky.social
Spencer Jones
@sjones6.bsky.social
Staff Software Engineer. AI Wannapreneur.

Medellín, Colombia 🇨🇴
“Two jobs ago I solved this problem in a repo that’s probably been deleted by now” 🤪
the hardest problem in computer science is where did I save the code I already wrote to solve this problem
December 26, 2024 at 1:09 AM
This is what I’m asking myself in 2025.

I waver between spending an hour or two each night to make progress and being a PlayStation 🤪
December 23, 2024 at 6:41 PM
Hey @qianli.dev , excited about you’re working on with DBOS.

But question for AI agents and workflows.

If the sequence of steps is not deterministic—ie, the agent decides what needs to be done based on input and decision engine—is that ok with DBOS?
December 22, 2024 at 1:45 AM
Perfect.
From 2017 but relevant every year:

“I don’t see any reason to be optimistic. What will the new year bring?”

“I think it will bring flowers.”

“Really? Why?”

“Because I’m planting flowers.”

(By the Spanish cartoonist José María Nieto.)
December 22, 2024 at 12:57 AM
A good read on strategy and how to avoid just sprinkling AI on stuff.

exec-edu.parlance-labs.com/02_asset_ai_...
asset_ai_strategy_toolkit
exec-edu.parlance-labs.com
December 21, 2024 at 7:54 PM
I have 3 small kids, which means a lot of home cooked meals with screaming not meaningful conversations.

One day I’ll have a good nights sleep and slow morning 😅
True luxury is found in the simplest moments.
December 18, 2024 at 11:56 PM
Listening to call recordings in 1.5x makes people sounds really smart.

They think so fast at 1.5x.
December 18, 2024 at 6:05 PM
December 18, 2024 at 12:20 PM
I’ve been considering building an open source prompting framework-ish project:

- defined prompts in yaml/json
- eval cases in yaml/json
- CLI to run evals
- CLI that publishes prompt SDKs in different languages

The idea here is that bringing together prompts and evals in a structured way …
December 17, 2024 at 10:50 PM
I stood up an instance of LiteLLM in Render.

I like the idea of what LiteLLM is doing … unified interface for lots of models (OpenAI, Anthropocene, etc), token and spend tracking, team management.

But, it also feels very immature as a product.
December 17, 2024 at 10:45 PM
I’ve been learning Spanish for the past 18 months.

Stuff I’ve been done: private tutor in person 2-3x week, online sessions, watching shows in Spanish, reading books in Spanish, in-person events and casual hangouts.

First time: conversation with advance voice mode in ChatGPT 🤯
December 16, 2024 at 1:46 AM
I studied the theology and the Bible, including Greek, Hebrew and a couple other languages. History from the second temple period.

Wrote a 200 page thesis on ancient translation practices from Hebrew to Greek.

Today, I’m a mix of software engineer and product manager.
Did you study something that has nothing to do with what you do for a living today?

I'll start 👉 I went to film school 🎥🎬

Your turn 👇
December 15, 2024 at 10:27 PM
Lol, best interaction on LinkedIn today 🤪
December 13, 2024 at 1:45 PM
The internet: hey look, AI can build a whole app in like minutes with just a simple prompt

Me (a staff engineer): spends 30 minutes trying to figure out z-index issues on a tooltip
December 6, 2024 at 4:02 PM
@kiwicopple.bsky.social , what do you think about adding static site hosting to Supabase?

Something like a vite/react setup.

Then you could do a full-stack Supabase app, auth, DB, storage, queues, functions, everything you need.
December 5, 2024 at 11:38 PM
Stoked about this.

Supabase eating the stack.
the big one: @supabase.com Queues ⚡️

Manage long-running tasks in Postgres. Features:

⬩ RLS support
⬩ Works with the Supabase libs
⬩ Built-in monitoring/observability
⬩ Message archiving
⬩ API parity with AWS SQS and RSMQ

youtu.be/UEwfaElBnZk
We added the missing piece! 🧩
YouTube video by Supabase
youtu.be
December 5, 2024 at 11:37 PM
Been tracking updates on some AI agent frameworks, here are a couple that I’ve seen.

Will add to the thread as a I learn about more.
December 5, 2024 at 10:34 PM
@supabase.com Launch Week, Medellin edition
December 3, 2024 at 11:34 PM
Honey wake up, a new agent framework from pydantic dropped.

ai.pydantic.dev
Introduction
Agent Framework / shim to use Pydantic with LLMs
ai.pydantic.dev
December 3, 2024 at 12:32 AM
Generating code from openapi specs is 😘😘

Setup fastapi to spit out an openapi spec, then generate a strongly-typed client in typescript to pull into a next.js app.

Great type safety across languages, minimal config.
December 1, 2024 at 11:07 PM
Feeling very proud… got a python setup configured with fastapi.

Connected together my next frontend, to Python with Supabase auth and storage.

Can upload a file via frontend, download and analyze it in fastapi layer.
December 1, 2024 at 1:21 PM
I’m trying to learn more python … have spent most of my last 7 years in Node/TypeScript land.

Man, I get wrecked with Python environments every time.

What is this!? 🤪
December 1, 2024 at 2:21 AM
Dónde están mis Supa amigos en Medellín?

Voy a asistir el Launch Week 13 de @supabase.com

lu.ma/pjch77cn?tk=...
Medellin, Colombia - LW13 Supabase Meetup · Luma
Supabase llega por segunda vez a Medellin ! Únete y celebra con nosotros! Este evento será una excelente oportunidad para hacer networking, compartir…
lu.ma
November 30, 2024 at 1:04 AM
Worst landing page ever.

Something related to AI, data analysis, but more like finding stories in data rather than making you dig for it

There's a bunch of apps that do text-to-sql or "chat with dataset" that are writing python or R and executing it in a sandbox.

Shooting for something different
November 29, 2024 at 2:08 AM
Reposted by Spencer Jones
It's time y'all
go.bsky.app/EWR2hu1

raise your hand in the comments to be added
November 28, 2024 at 8:55 AM