Andrew Moorcroft

Hey @davidfowl / @shanselman - what .Net project excites you for 26?

@brankopetric00 Take a look at fck-nat.dev/stable/ it basically does exactly the same thing as NAT gateway for just the price of a tiny EC2 instance. Saved us a tonne when we deployed it

Our NAT Gateway cost more than our entire application infrastructure. $4,800/month for routing. The discovery: - AWS bill jumped from $12k to $17k - Investigated line by line - NAT Gateway: $4,800/month - Wait, what? What NAT Gateway does: - Allows private subnets to access internet - $0.045 per GB processed - $0.045 per hour running Our usage: - 3 NAT Gateways (high availability) - 89 TB data processed monthly - Mostly outbound API calls Where the traffic came from: - Microservices calling external APIs - Pulling Docker images from Docker Hub - Downloading packages during builds - CloudWatch logs - S3 uploads (could use Gateway Endpoint) Optimizations: 1. VPC Endpoints (free data transfer): - Added S3 Gateway Endpoint - Added DynamoDB Gateway Endpoint - Saved 34 TB/month 2. ECR instead of Docker Hub: - Moved all images to ECR - No NAT Gateway charges for pulls - Saved 22 TB/month 3. Consolidated API calls: - Batched requests - Added caching layer - Reduced external calls by 60% - Saved 18 TB/month 4. Reduced NAT Gateways: - Went from 3 to 1 - Acceptable risk for non-production - Saved $720/month in hourly charges New monthly cost: - 15 TB through NAT Gateway - $675 data processing - $240 hourly (1 gateway) - Total: $915/month Savings: $3,885 monthly ($46,620 annually) Lesson: Data transfer is expensive. VPC Endpoints are free. Audit your traffic before scaling NAT Gateways. Architecture decisions have billing implications.

DeepSeek just proved the 'worthless' GPT wrapper startups are actually the ones with real moats. A week ago, nothing was more LOW status than being a 'GPT wrapper' startup. But I think we're learning that's DEAD wrong. Turns out they were just early to the only game that matters While DeepSeek, Meta, Anthropic and Microsoft battle over benchmark scores, these 'wrapper' companies have been quietly building the only moat that matters: interface loyalty. Because of how big owning the LLM is for national defense and the economy, the next breakthrough model is always 2 weeks away. DeepSeek launches today, someone else drops a better one tomorrow. But getting millions of people to make your product part of their daily workflow - that's the real barrier to entry. ChatGPT didn't win because it had the best model. It won because it was dead simple to use. And I think it has staying power because of that. This is why all those AI startups we dismissed are actually positioned to win. They're not competing on model performance - they're competing on being the default way humans interact with AI. Frontier models are becoming commodities. User habits aren't. While everyone obsesses over the next architecture breakthrough, the real game is being played in the interface layer. The moat isn't in the model - it's in being the tool people reach for without thinking. Technology advantage is temporary. Interface lock-in is forever. Keep shipping those wrappers, my friends.

@davidfowl Ah sorry I missed that 🫠🤣

@amoorcroft1 I'm talking about personal projects 😅

The results of this post are interesting. Very often the extra “deployment engine” on to of the core pipeline isn’t required (I understand some deployment engines have value add features). The diversity in this space is really amazing. #devops

@davidfowl Makes sense. Yeah I’ve just used GitLab self hosted for years and have always found it very powerful, has most of what I need out of the box. There’s basic templated CI pipelines you can use or you can build more complicated if needed. Assume at MS you’re restricted to GitHub!

@amoorcroft1 That's because my source would be on github. If my source were on Azure DevOps, I would use DevOps pipelines (in fact, those are more powerful than github actions).

@davidfowl Can you replace docker compose with it (by pulling in images as opposed to local projects)?

This year we’re taking aspire to the next level… #dotnet #aspire

@davidfowl GitLab for source control and CI. Pipelines set up to build and push to AWS ECR. Then a templated Cloud Formation stack to take that and update AWS ECS

What system are you using to automate deployment?

@levelsio @adamwathan Not really a tool though is it?

@adamwathan Property tax

What tool that your company spends at least $1000/yr on do you hate the most?

@idlesband what a night, surely the best UK band currently touring? Spent the last 45 in the pit 🥵

🚀🚀🚀🚀🚀🚀🚀🚀 #dotnet

@Tim_Burgess Go on then 😆

If this tweet gets 50 likes I’ll play you my official Euros Final song 😉

@bretajohnson @davidfowl @JamesNK @Microsoft Microservice architecture over mainly HTTP. The local dev experience involved running ~10 sln’s (even moved to Rider from VS to help with this). Aspire has enabled us to just run one solution and provision all the dependencies, as well as db / redis / rabbit, need to dig further

@amoorcroft1 @davidfowl @JamesNK @Microsoft What's the project type that you added Aspire to? Was there anything specific that you especially liked - or thought could be improved - in the experience?

we added .Net Aspire to an existing project today, was surprisingly easy to integrate, and get instant actionable feedback. Shout out to @davidfowl / @JamesNK and others at @Microsoft for the hard work

@Camelboy2Meow @davidfowl @JamesNK @Microsoft @reubenbond yeah sorry, wasn't sure who all the big hitters were!

@kieranmch Sorry assumed you meant ChatGPT.. Maybe don’t look up Dr Gupta AI if you think this is bad btw!

@amoorcroft1 Llama is a GPT, and putting a lot of effort in is not the same as regulatory approval

Feeling quite uneasy about people marketing their GPT wrappers as therapists without any testing or trials or clinical validation. I think regulators will have to come down on this stuff.

@dylanbeattie @dgkimpton An example being @browserstack. Every now and again I’d like to run some manual cross device tests. I’d happily pay for an hour or day access, but I’ve found a way to get it free for a few mins at a time using temp emails. The only paid plan locks me in for monthly recurring

@dylanbeattie @dgkimpton From experience it’s usually the licensing. 95% of the time you’re having to buy in to a recurring subscription for something you’d happily pay for to use for a week or two

Memories of Kiko Macheda with that goal #hojlund #manutd