Shaun Davis

@zuess05 In critical software, code generation isn't the bottleneck. Verification is. Messy code is what turns a 3-hour change into a 3-day exercise in proving it's safe.

Senior developers are currently having a massive existential crisis because Claude writes "messy code" A junior just used Claude to ship an entire feature in 2 hours. Meanwhile, the Senior is still spending 3 days reviewing code. When will y'all realize that literally nobody cares if the code is "messy"?

Alex wrote an MCP server for Sprites, and Kurt is in his feels about it. Still, even if you're using MCP to do it, "On a new Sprite, do XYZ" is a cool feature for an agent to have. fly.io/blog/unfortuna…

@rvivek Code generation is cheap, but validation is expensive.

AI is supposed to make us more productive. Yet, every single person I meet is working way harder than ever before. Explain the conundrum.

@mattvanswol I've been fighting with similar questions the last couple of years... The best i've come up with is: Social skills, critical thinking and being adaptable...

I could use some help. My kids (middle/high school) are already thinking about college and life after graduation. Looking 5–10 years out, it’s hard to find a path that doesn’t look like AI could replace it. If you were them, what would you study or do after high school?

@BrianNorgard Yep...

Almost every smart person I know who works in technology is experiencing extreme levels of anxiety. As if the rug is about to be pulled once and for all.

@rasmalai The direction we are going isn't well suited for capitalism. We are going to have to transition to something new. What that will look like is the big open question.

@nicbarkeragain To be fair, I never particularly enjoyed reviewing non-llm generated code either.

"AI will generate all the code, and senior programmers will review it" is basically a mantra at this point. I'm fortunate enough to know a number of top tier programmers personally, and precisely 0% of them want to spend their time reviewing AI generated code (myself included)

@xoaanya We are moving toward JIT software. Security will be the hard technical limit. Systems thinking is the new leverage.

Anthropic’s CEO says software engineering could be obsolete within 12 months. Agree or overhyped AI doom take?

@Coinvo Semantics. It's illogical to think that Ai will replace software, but that says nothing about the role of humans creating it.

BRAKING: Nvidia CEO Jensen Huang calls it "illogical" to think AI will replace software and related tools.

@AdvicebyAimar @paulg I don’t think it’s generally a good idea for tech leaders to insert themselves into political wars. Their leverage comes from building useful, trusted systems, not from picking sides in volatile political conflicts.

@paulg its part of the reason many people love tech but hate tech culture. most tech leaders have zero spine and bend towards whatever narrative favours them the most. tech has such a huge influence on all of this, its weird non of them are taking a lead in this.

He's right you know. How bad do things have to get before you say something?

This has been said a thousand times before, but allow me to add my own voice: the era of humans writing code is over. Disturbing for those of us who identify as SWEs, but no less true. That's not to say SWEs don't have work to do, but writing syntax directly is not it.

@hayesdev_ Very cool, but this seems like a bit of a security nightmare.

Cool follow up post about the design behind the Sprites agent sandboxes from Fly dot IO (from Thomas Ptacek himself :-)) fly.io/blog/design-an…

We made a thing called Sprites that lets you run AI code safely. Watch Chris create sandboxes in seconds, let agents install packages freely, then restore to checkpoints when things go sideways. 🔗 ➡️ 🧵

Agent Sandboxes with Sprites.Dev from @flydotio Shot a quick video walkthrough of my first Sprite sandboxes this morning youtube.com/watch?v=KLP6RB…

@sama Sora 2 is both incredible and deeply disconcerting. The creative potential is undeniable, but the level of realism is beginning to edge into a form of manipulation, which i'm becoming increasingly less cool with. How do we prevent this from becoming a huge problem?

Introducing Phoenix.new: An AI-powered Phoenix app builder running on Fly Machines. Build Elixir apps with an AI assistant that has its own headless Chrome to test what it builds! 🏗️ 🔗 ➡ 🧵

@acodechef Really just comes down to what your doing.

@davissp14 Yaaas, my intuition is working. The other half of my brain was telling me you’d want those writes on disk as quick as possible. What about DIRECT_IO usage?

Database folks: does batching writes to your WAL file, and then flushing when the upper layers commit the transaction, make sense?”

@acodechef Haha. Well, there are some circumstances where it could make sense. Like, high durability systems like Etcd will write each change to WAL and persist it to stable storage before replicating it.