Tessera
Folgen

Luke Hedger

@lukehedger.dev

Software engineer. Building things for lego.com. Distributed systems, serverless, event sourcing, figuring out ai. Wrote Serverless Development on AWS (O’Reilly) https://lukehedger.dev/ https://lsh.pckt.blog/ 📍 London

576 Posts407 Follower99 Folgt
Luke Hedger@lukehedger.dev · 6d ago

it just needs one absolutely killer app and the strong network effect/portability will kick in web2 didn’t form overnight - it’s been a slow decay since facey-b keep leading with data ownership but make the app soooo good you _have_ to use it and stop being “this, but on atproto”

dame@dame.is · 6d ago

hard pill to swallow for atproto developers/creators: 99% of potential users either do not care about, value, understand, or want to “own their own data” i see so many apps/projects lead with this “value prop”, but it’s not an important or meaningful consideration for most people

Luke Hedger@lukehedger.dev · 6d ago

atproto does feel familiar to web3 but I feel as if the community has already decided “we’re not that” and will risk not learning lessons from the failure of that movement lsh.pckt.blog/is-atproto-b...

is atproto blockchain without the coins?

“web3 but good”?

lsh.pckt.blog

dame@dame.is · 6d ago

those of us who were embedded deeply in the earlier “web3” movement learned this lesson awhile ago, and after building things in the atmosphere everyday for the past 2 years i can confidently say that “data ownership” is still not worth focusing on in most cases

conputer dipshit@davidcrespo.bsky.social · 2w ago

“you should do a breakthrough” is also a great illustration that when LLMs actually get good at something, it doesn’t mean they *can* do it with careful prompting, it means they can just do it. we’ve seen task after task in software engineering go from needing careful prompting to “ok now do it”

conputer dipshit@davidcrespo.bsky.social · 3mo ago

it’s a very good piece. my thought as someone who spends 40+ hours a week looking for and finding LLM-shaped problems is that you don’t have to push it. things are hard and then a new model or tool comes out and they suddenly become very easy nymag.com/intelligence...

This is a recurring theme when you try out new AI tools. You recognize that there’s a lot that might be done with them, but not much comes to you. You see this in the rise of AI coding tools, which you find extraordinarily impressive as you use them to … make yourself another … news reader? Notes app? Personal website, again? The tool you made for ripping and labeling files out of a popular music service gave you a slight, Napster-ish illicit thrill and actually worked, but your use case — putting them on an old MP3 player to run with — was an aspirational mirage, a task invented to have one. You ask if this is a failure of imagination, a personality flaw, a matter of creativity or practice, and if you’re just sitting in front of a piano that you don’t yet know how to play (some of your programmer friends don’t have this problem, they say, and are swimming in bespoke apps of their own creation). You also ask, perhaps, if your inability to find little software-shaped problems to solve in your life is rooted in psychology and related to an awareness that, from the outside, your entire job looks an awful lot like someone else’s software-shaped problem (the first thing public LLMs could do, after all, was produce novel, readable piles of text).
Luke Hedger@lukehedger.dev · 3w ago

if your traffic spikes to unpredictable levels, where you have no idea about how many servers you need, choose serverless the barrier to entry is so low now and the footguns harder to set off with the right framework

Luke Hedger@lukehedger.dev · 3w ago

if you’re running exclusively on ecs/ec2/eks you probably shouldn’t be renting compute from aws if you’re running compute on lambda but don’t have extremely variable traffic you probably shouldn’t be the cloud is rarely going to be a good fit nowadays