Ruben Cuevas

Experimenting with AI agents | Sr SDE at Amazon

[Ruben Cuevas @rubencu]

@simpsoka in codex app would like to have the name of the worktree associated to the thread easily visible at a glance (or hovering over)

[Ruben Cuevas @rubencu]

I haven't found a replacement for an IDE. I do all my design and coding through CLI / Codex App, but I always keep projects open in an IDE (mainly Kiro) to:

1) review all outputs
2) get a visual understanding of how the project is structured and where my changes are located. I iterate a lot with agents, asking questions like "why are you touching this module vs this other one"
3) foster serendipity. Navigate the project, explore the surroundings of my changes. I find this also important to get familiar with the codebase, and help discover tech debt and refactor opportunities

I find the IDE / no IDE discussion a bit noisy. If you're working alone on your own repos, yes you can thrive without an IDE. It changes a lot if you're submitting PRs to open source projects. You need to convince maintainers that your PR is in the best possible form to get merged, and I find it difficult to defend that I'm touching the right places without getting this "visual understanding".

Yes, you can also chat with agents to get the visual understanding, but then I refer to point 1) review all outputs. I still need to confirm myself.

[jack friks @jackfriks]

the only problem working without and IDE is if you work with others who do use an IDE then they will hate you cause no one who looks at the diff would push that +20,000 lines added commit to main.

[Ruben Cuevas @rubencu]

isn’t the AI frontier models gatekeeping making the Bitcoin narrative stronger?

[Ruben Cuevas @rubencu]

it’s a great time for generalists / people who went too deep into their hobbies

[Matthew Botvinick @mattbotvinick]

We're building a team at @Anthropic focusing on AI and the rule of law. We've made our first hires, and are now opening up a new research engineer role. We're looking for people with advanced technical skills, including AI/deep learning/NLP, full-stack development and data science, paired with training or experience in law, government, political science, or a related field. If this is you, or a friend, please get in touch.
https://t.co/uZQEZC5lck

[Ruben Cuevas @rubencu]

@kr0der also not all necks can deal with 2+ monitors

[Ruben Cuevas @rubencu]

I panic bought 3 MacBook Pro M5 Max 128gb RAM that I found at the old price

was thinking on keeping one, but also I might be able to do some cool stuff with the 3 like setting up a cluster with @exolabs

any suggestions?

[Ruben Cuevas @rubencu]

@antoniogm buenos centollos

[Ruben Cuevas @rubencu]

I bought an iPad Air some days ago but never end up pulling the trigger on the maxed out M5 Macbook Pro. Sad..

[Marques Brownlee @MKBHD]

Memory price hikes have finally started hitting Apple. Today, they raised some base prices:

Macbook Neo: $599 -> $699
Macbook Air: $1099 -> $1299
Macbook Pro: $1699 -> $1999
Mac Studio: $1999 -> $2499
iPad Air $599 - $749
iPad Pro: $999 -> $1199

[Ruben Cuevas @rubencu]

The jump to the third paradigm is going to be harder.

First and second implied you iterated with the agent interactively, like an extension of you, even as agents kept gaining capabilities and working more autonomously

Third paradigm is treating agents like an actual entity 100% detached from yourself. It’s not you with superpowers anymore. It’s someone else.

The mental model shift is way harder. This transition will be slower as people are already happy on the first and second paradigms

[Andrej Karpathy @karpathy]

This is a new paradigm for interacting with Claude that is significantly more "inline" with all the other human activity org-wide. Once you do all of the under the hood engineering work to make this "just work" (e.g. across tools, integrations, compute environments, memory, security, etc.), Claude basically joins the team in a seamless way - you can talk to it as you would talk to a person and it can help with a very large variety of workloads.

Imo this is the 3rd major redesign of LLM UIUX. The first paradigm was that the LLM is a website you go to, the second was that it is an app you download to your computer. This third one is that it is a self-contained, persistent, asynchronous entity with org-wide tools and context, working alongside teams of humans. It really takes a while to wrap your head around it, but it works and it is awesome.

[Ruben Cuevas @rubencu]

@MattSalsamendi @leerob it’s just that working remotely requires a set of skills that are usually learned by working in some high stakes corporate environment in person. Requires previous experience

[Ruben Cuevas @rubencu]

@Semper_Viventem same for bugs and mistakes

[Ruben Cuevas @rubencu]

@jessegenet lack of empathy. “I’d rather risk cancer” until it’s your turn

[Ruben Cuevas @rubencu]

@benhylak can I just have confidence that I’ll assign you some work and you’ll get it done in a reasonable way? single criteria for promo to next level

[Ruben Cuevas @rubencu]

Especially if the transition happens within the same company, better to learn the culture in person first

[Ruben Cuevas @rubencu]

@leerob I’d say having experienced in-office work is a prerequisite for later successful remote work

[Ruben Cuevas @rubencu]

100% agree. Years of exp also matter. Hard for remote work to be net positive for new grads in the long term

[Lee Robinson @leerob]

Both of these things are true:

1. Remote work is hard and can be a bad fit for many
2. Office culture can be great and more productive

Rather than ignoring nuance, just don't hire the person who needs office culture remotely, or vice versa.

[Ruben Cuevas @rubencu]

@chrisalbon ChatGPT Pro

[Ruben Cuevas @rubencu]

@marckohlbrugge @cursor_ai I’m super late on this but how did we end up figuring out that Kanban is the way to manage agents?

[Ruben Cuevas @rubencu]

Noticing great memory improvements in ChatGPT and Grok vs months ago.

Memory kept evolving but doesn't get as much attention as other capabilities. It's something you just notice one detail at a time.

[Ruben Cuevas @rubencu]

Codex has /clear in the CLI but not in the app. Why?