Tom MacWright
@macwright.com
about 2 months ago i hope this is not subtweeting my prs 😂
💬 1
♻️ 0
❤️ 2
i hope this is not subtweeting my prs 😂
Claude Sonnet 4.6 is now available in Netlify AI Gateway and Agent Runners.
If you’re using the Anthropic SDK, switch to claude-sonnet-4-6 and go.
www.netlify.com/changelog/cl... Didn't try out Rolldown yet?
You should as it entered RC recently and is 10-30x faster while being compatible with Rollup's excellent API.
rolldown.rs If you're a library author adding `@__NO_SIDE_EFFECTS__` to your code, know that your users will only see the full benefit with a bundler that understands it before splitting.
This matters for libraries like shadcn, Radix, or any component library using the annotation. Your route-level bundles can actually get smaller from it!
But @rolldown.rs handles this. It tracks which functions are marked `@__NO_SIDE_EFFECTS__` across your entire app. When it sees a call to one of those function, even from a different file, it knows the call can be safely dropped if the result is unused. If your bundler only checks the annotation during minification, it's too late The minifier only sees one file at a time Which means it sees the call to the function but has no idea that this function was marked as side-effect-free before 🚨 The annotation does nothing exactly where it matters most!
Most apps use code splitting: Your code gets split into multiple files (chunks) to load less unnecessary JavaScript. When that happens, the annotated function might end up in one file and the call to it in another.
Did you know about the `/*@__NO_SIDE_EFFECTS__*/` annotation? It tells bundlers that a function has no side effects so every call to it can be tree-shaken if the result is unused. Most bundlers and minifiers support it today. But there's a nuance most people miss 👇
Clip from our internal showcase: Sahn Kim-Kuo took a prospect call recording → fed it into Claude → used Netlify Agent Runners + brand guidelines → built a custom webpage follow-up.
First one took ~6 hours. Now it’s a ~30 min rebrandable template.
Full showcase: www.youtube.com/watch?v=ML03... You also bet we blogged about it:
blog.railway.com/p/network-fl... Just GA'ed View private network traffic directly on the canvas. True sight, but for your network. Available today. Watch Phin, Platform Engineer, talk and demo the feature below.
"Vibe coding" isn't a meme. It's how people are actually shipping now.
Describe what you want. AI writes the code. You iterate, deploy, and learn along the way. Here's the full breakdown of what it is and where it breaks down.
netlify.com/guides/what-is-vibe-coding Coming in tomorrow's release: granular per-tool permission settings for Zed's Agent. Configure regex patterns for any tool to always allow, deny, or ask. Your agent. Your rules.
this whole discussion is about vibes, the initial question is why people associate ai with crypto and don't like it, that's the thing we're talking about, that is the topic
when people get laid off from their jobs and the employer states "it's partly because of ai" the people getting laid off are going to blame ai because their employer blamed ai, there is no need for 3D chess here
my partner's electric bill has gone up 250% and two friends have been laid off, yes
for crypto, lots and lots of people losing money, especially middle class people and Black people were the ones who suffered from crypto scams. for ai, people losing work, bosses demanding more, and right now, paying higher electricity bills. that has been extremely polarizing.
where previous technologies have been oppositional they've mostly been that in opposition to companies, not individuals, never touching the level of interpersonal warfare that these technologies seem to have at their absolute core.
crypto was an oppositional technology to central banks, etc, reinforced by constantly telling people that they'll 'stay poor' and so on. it was built on resentments. ai has done the same: providing robot artists to demean the work of real ones. nearly gloating about job destruction.
which may be true! it's a more real technology. okay, after having thought about it I have a much clearer response to your initial prompt: because they are technologies with clearly defined losers, which are almost defined by how they define losers
i totally agree that llms are a more real technology with actual adoption, but i can log into linkedin and provide you with an endless stream of stupid claims, or look at openai's webpage for the sentence "It may be difficult to know what role money will play in a post-AGI world."
at the height of crypto mania, people said that mortgages would be on the blockchain, websites would be all on web3, and fiat currency was on its way out. i can't find a parallel to any of the other trends on that list other than llms, which people imagine as replacing capitalism itself.
imho because - they're both post-2016 largely right-wing cultural movements with a lot of uniquely odious people at the helm - unlike the rest of the list, they've both had dramatic personal impact on people's finances, mostly negative - both are much more totalizing than the rest