dilski 🦆

@UsagiKaede1 There's got to be some caveat that prepared spells move onto the stack, and therefore aren't on the battlefield - and so not available to cast again

You unprepare once the spell resolves so you can just pay 5 blue and cast 5 copies of Ancestral Recall.

@sainsburys should this substitution not be 3x the 8 pack rather than 2x?

@charliermarsh @davegaeddert Rad

@davegaeddert I also built this which generalizes the pattern though haven't mentioned it publicly anywhere: uvx.sh

@charliermarsh any specific tips for writing an install script that checks for/installs uv as part of it? I remember you tweeting about one but can't find it. Just curious if there is a "standard" for this by now

had dinner with wife at a mexican restaurant last night looked at menu, they were trying to raise prices from $18 to $24 for her favorite entree wife was like "i think we can have claude make this" told waitress trying to gouge us they done 1 week sprint cloned & replaced cochinita pibil and carnitas restaurant manager freaks out "how do we solve this" this is going to happen so much in 2026

@gothkuro @khubiez Discord file sharing is not designed for this, so it's not as secure or durable as something which is designed for that. Discord attachments are served through their CDN - anyone with the link can access without any authentication or authorisation, even if you delete the file.

@khubiez This is so funny to me because like why not?

oh my god do not do this

@theburningmonk According the the SAM source, there should be a capacity provider config on the AWS::Lambda::Function and a AWS::Lambda::CapacityProvider. There are links to where it should be in the cloudformation schema docs but they're dead links. docs.aws.amazon.com/AWSCloudFormat…

Looks like you have to use SAM's AWS::Serverless::Function and AWS::Serverless::CapacityProvider types to create a Lambda function with managed instances for now. Am I missing something here? 🤔

mm.. did AWS release Lambda Managed Instances without CloudFormation support? The AWS::Lambda::Function type doesn't let you choose the managed instance option, but the CreateFunction API does. And there doesn't appear to be a CloudFormation type for Lambda capacity providers.

@__steele @theburningmonk #diff-f99abdf19eef43403413f6070c37d7e085b6ab146459f3590c8ca107f2bdc2edR58" target="_blank" rel="nofollow noopener">github.com/aws/serverless…

@theburningmonk What do these resolve to? Doesn’t the serverless transform resolve to “native” CloudFormation types?

@astuyve The testing library is lovely too! Options between running locally or incoming functions in the cloud!

Lambda provides an SDK built into the `context` object which lets you do things like register a callback to wait for an async endpoint to respond. You can just make the function sleep, or poll and wait for some state to change (eg: wake up, try an endpoint sleep again)

NEW: AWS Lambda Durable Functions are LIVE. If you ever wanted to build step functions expressed as code, this launch is for you!

@QuinnyPig @awscloud x.com/dilski/status/…

Welcome to the #awsreinvent keynote thread. Apparently they're livestreaming the keynote via Fortnite this year; either @awscloud or I am very much mistaken about what "Fortnite" is and I honestly don't know which it is.

Watching the #AWSreInvent keynote in Fortnite as god intended

@astuyve This is pretty cool. I'm guessing you can use this to give lambda GPU?

NEW: Lambda FINALLY adds multiconcurrency support with Lambda Managed Instances! Sandboxes process multiple concurrent requests when deployed into EC2 clusters managed by AWS. You pick the instances, and you can set the max concurrency per sandbox. MEANING: No more cold starts!

@MarcJBrooker How do you go about managing schema? It's the 1 thing putting me off trying dsql instead of DDB for my serverless applications

What about the added complexity? Well, there isn't much - I created this database with a single API call and never have to operate it in any way. My code does have to use a DB client or ORM, which does add some work, but it's local and relatively easy.

Why use a database and not just put these 2000 rows in memory? Clearly it's not about scale. My laptop could fit 2000 rows of this data in memory about 10,000 times over. But this isn't a static dataset, it's being actively modified. This is where databases become important. 1/

@cezthesocialist @elliseccles @NiallHignett COTTAGES - Five fully modernised, traditional farm cottages – three on short-term lets and two forming part of a thriving holiday letting business

It has 3 homes on it which they plan to use as short-term lets and land, especially a huge piece of land like that, is a very good investment at the moment. They are not poor. And the inheritance tax he will have to pay for getting a farm for free is a much reduced rate and can be paid over 10 years.

PARENTS: please check your kid's candy this halloween - i just found an 18-month enterprise salesforce contract with AI features and a $100k early breakup fee inside this snickers bar

my hottest take is that conda is objectively better than uv 1. conda is short for anaconda, which is a type of snake. guess what, so is a python 2. used primarily by academia and they are famously smart people, so they must know something we dont 3. "conda activate" makes you feel like you're commanding a tank. "uv run" sounds like you're doing laundry 4. it takes 45 minutes to resolve dependencies which gives you time to smash another white monster and watch youtube at work

@astuyve As far as I'm aware the alternative is either an always-running collector instance (and then your lambdas all have to be in a vpc and have network access to the instance), or something custom like @donkersgood did (lucvandonkersgoed.com/2023/11/16/ser…)

@astuyve From my understanding the cold start is due partially to OTel, but a big chunk is additionally having to run the collector as part of the invocation - the x-ray equivalent (the x-ray daemon) runs as part of "the fabric of lambda" so doesn't impact invocations.

AWS X-Ray SDK and daemon have officially posted an EOL timeline and recommend migrating before Feb 27

They really do just chuck out this AI slop and expect you to vote Labour at the next election. No idea who their audience is, no real comms strategy across government and paving the way for Reform by patronising the electorate with this junk.

“We must all be free to debate openly and freely without fear - there can be no justification for political violence.”