Alphonse

@samgoodwin89 Yeah your right I guess a tunnel can work!

@Alphonste You can use tunnels or whitelist cloudflare’s IP range. Those are both less than ideal, that’s for sure.

Have been building resources for deploying websites to AWS and OH MY GOD CloudFront is hot garbage compared to Cloudflare. 10+ minutes to deploy vs ~1s. Do people really want to deploy websites to AWS?

@samgoodwin89 For cloudflare to connect to RDS inside a VPC, it must be accessible publicly or else hyperdrive can't connect to it. Maybe this doesn't matter practically but I like the idea of the database being isolated and I'm sure other aws users do too.

@Alphonste You can connect your DB to Hyperdrive securely AFAIK

@birdabo you think this is what a phd in ml does? Your actually retarded

Karpathy dropped an AI that does autonomous ML research. > picks architectures > tunes hyperparameters > commits code changes. > runs experiments in a loop > zero human input degrees and PhD are gonna be useless in 6 months.

@sircatstock @tobi Brother thinks shopify could've existed without canada almost free education + infra + free healthcare etc. He's retarded

@tobi Billionaire moron amplifying populist nonsense.

@tannerlinsley @Neolectron @EffectTS_ I've found ai to be actually incredible at writing it

It was basically a new language. It was hard enough to convince myself to learn TS, then convince everyone else I know to write types for it. Effect is another magnitude jump into proprietary land. I can't imagine onboarding a new dev or contributor to it, let alone trying to get AI to write in it.

I need @tannerlinsley to get into @EffectTS_ . How is it not already the case, a huge typesafety stan should use more types. You probably tried, how did you resist the urge to ship it ?

@vix_0ii @frankdedomiseur lmfao charles milliard

@frankdedomiseur Les péquistes s'étaient convaincus qu'il y aurait PSPP Premier Ministre et un référendum gagnant. Maintenant ils auront ni l'un ni l'autre. Votre prochain premier ministre :

Le Meltdown sur le sondage Léger est assez... comment dire ? Intense.

@real_Ivn @craigzLiszt What is there to innovate on? It perfectly does its job.

docusign could really just be a team of 1

@spacetime_db @NickVanWig wtf does that even mean, a call on pg?

PlanetScale got video calls working in Postgres! Pretty incredible, @NickVanWig! Never underestimate the power of databases. They posted the code too: github.com/nickvanw/PgVid…

So help me understand what you mean by runtime here. Usually in sst, to package a lambda, its pretty simple. Point to the handler file, and boom this gets packaged in a way I understand 0% about, but it gets packaged and ship to AWS. So with alchemy-effect, your plan is to remove this step? Don't you still have somehow to package the code, send it to aws somehow? But your point is that effect automatically tree shakes the package that gets sent?

@Alphonste The closure inside Effect.gen runs once per “Runtime environment”, eg Lambda Function. It is the entrypoint for your service. Both at plantime and runtime.

Here's all it takes to create a S3 Bucket, SQS Queue and forward Bucket Notifications to the Queue with `Stream.tapSink`. Alchemy-effect creates all the Resources, IAM Policies and Environment variables automatically. Plan is to have a release ready for usage this week.

@samgoodwin89 Okay I'm still quite new to Effect but I can see the vision, bro this is gonna be just incredible lmao can't wait to play with it

@Alphonste Another file brings all your Functions/Resorucs together into a Stack Effect. That Stack Effect is then used to generate and apply a Plan. That's where we provide all the IaC layers since they are only needed when deploying and can be omitted from the Function code entirely.

@samgoodwin89 Ahhh shit smart, and then when you merge the layers and provide them, they concurently get created based on their requirements I imagine?

@Alphonste It's a Resource Effect. It doesn't create it when you call `yield*`, it adds it to your Plan. This same code is then run at runtime to execute the Runtime Effects. All the IaC code is decoupled as Layers, so your runtime bundle is tiny.

@samgoodwin89 So you would like people to use alchemy-effect or alchemy?

@Alphonste Effect code should just slot in to this. If it doesn’t then we’ve failed.

I've been searching for the perfect Cloud primitive since I think 2019 and this is the closest I've ever felt. I've always seen Cloud programs as "two-phase" programs: 1. "Plan" code runs first to define "what" Resources to deploy. 2. "Runtime" code is deployed to the cloud where it executes the business logic by interacting with the Resources. But I never understood how to have these Phases interact in a sound, composable and type-safe way. After learning how to use Effect Layers, I am now imagining a multi-phase Layer concept where a "Plan" phase runs before (and envelopes) a "Runtime" phase. Shape: `Plan(Runtime(Handler)) ` For lack of a better name, we are calling them Constructs. Not just as an homage to our buddy, the AWS CDK, but also because of how a Construct is a general concept for building blocks. Does it make sense for a Construct == Two-phase Layer? I think it does, maybe (lol). The pattern goes as follows: 1. define a Construct.Tag (like a Context.Tag, but Phase-aware) 2. Phase.plan comes first and is where Resources are declared, e.g. `yield* Bucket(..)` 3. Embedded within the Plan, you can run Phase.runtime(..) and interact with Resources declared in the Plan phase. E.g. gain access to a generated Worker ID, Queue URL, etc. 4. Binding requirements (e.g. GetObject) are inferred from and attached to the Runtime Effect which is then absorbed into the Plan Effect. 5. Compose these together into 1..* entry-points (programs) and deploy them as a Stack. This pattern (alone) gives us: 1. Components containing both Resources and Runtime logic (hallelujah!). 2. IAM Policy inference and encapsulation (Bindings are enforced but don't leak out into consumers of a Construct) 3. Scriptability like Alchemy/Pulumi (conditionally create resources or tune properties based on stage or other parameters) This is feeling right to me, but this design problem has been a bottomless pit for me, so I always wonder if there's an even better way. Maybe it can be done solely with Layers and without introducing a new "Construct" idea?

@samgoodwin89 Real question, why not focus on AWS? Not an attack, just curious.

Cloudflare has a nasty habit of rolling out breaking changes to their API. We've been broken twice in two weeks by un-announced and un-documented changes. If you broke prod like this in AWS everyone would be fired.

@JannikWempe Yes bro I think its because of their other project open code. But SST is kinda best in class, open code has a bunch of competitor I don't really get it but yeah maybe it wasnt making enough money.

It’s sad… SST feels kind of dead. There is still no node 24 support 😢

@tokihossain @garrytan Good. Lets keep them here.

we have so many incredible builders in canada. even @garrytan is canadian not looking good

@beffjezos I think we can trace this back to some event in 2024, I wonder what it could be...

The adversaries' Cold War plan was always to sow cultural division within the Western world to divide and conquer us. Defection of commonwealth countries to the CCP was always their plan. IMO isolationism is not a good idea. America must maintain strong internal coherence while maintaining strong alliances. Anything else effectively weakens our position.

@ugbahisioma @tannerlinsley @tan_stack Yes ive faced this issue before. Give more memory to your lambdas. Go to the max its like 3048 or something

@tannerlinsley @tan_stack Is there a way to control this? Either in the app code or on netlify side?

I noticed that my @tan_stack start app/site if left for a while on a tab, and the you come back to it and try to navigate to another page, it takes a lil while.. like the server “slept” or something, but after you refresh it’s back to normal speed @tannerlinsley, is this normal?