Emil Privér
@priver.dev
Founding engineer and Part-Owner at http://enad.io | Love rust and go
I also don't think you get any more efficiency in organizations as well, often has companies made organizations structure changes that allow the developers to be more efficient but claim it on AI to make stakeholders and investors happy because the company is now "innovated"
January 11, 2026 at 1:49 PM
I also don't think you get any more efficiency in organizations as well, often has companies made organizations structure changes that allow the developers to be more efficient but claim it on AI to make stakeholders and investors happy because the company is now "innovated"
AI won't magically help you be able to ship software
January 11, 2026 at 1:49 PM
AI won't magically help you be able to ship software
It's equal important today that you understand & can reason about code and software as before the big rise of LLMs. If you don't understand this and rely on AI to write code for you will you ship even worse software then if you wrote it yourself without AI.
January 11, 2026 at 1:49 PM
It's equal important today that you understand & can reason about code and software as before the big rise of LLMs. If you don't understand this and rely on AI to write code for you will you ship even worse software then if you wrote it yourself without AI.
I don't think programming or building software have changed anything since AI have became a thing some years ago.
January 11, 2026 at 1:49 PM
I don't think programming or building software have changed anything since AI have became a thing some years ago.
We decided to use go for the "contracts" as we call them due to that it's faster and works good when everything is written in go
The idea is to use the jsonschemas/avro as a base for the events and then write code to convert jsonschemas into go, rust and python types
The idea is to use the jsonschemas/avro as a base for the events and then write code to convert jsonschemas into go, rust and python types
January 9, 2026 at 9:12 AM
We decided to use go for the "contracts" as we call them due to that it's faster and works good when everything is written in go
The idea is to use the jsonschemas/avro as a base for the events and then write code to convert jsonschemas into go, rust and python types
The idea is to use the jsonschemas/avro as a base for the events and then write code to convert jsonschemas into go, rust and python types
We have 1 package we use to validate the events we send within our system and this is currently written in go, the problem with this is that we get limited on writing applications only in go and we're have projects where rust and python both works better.
January 9, 2026 at 9:12 AM
We have 1 package we use to validate the events we send within our system and this is currently written in go, the problem with this is that we get limited on writing applications only in go and we're have projects where rust and python both works better.
Currently working on migrating away from a package written in go to jsonschema or Apache Avro.
January 9, 2026 at 9:12 AM
Currently working on migrating away from a package written in go to jsonschema or Apache Avro.
://
Lets hope it het sorted out good for @tailwindcss.com it's a phenomenal good product
TLDR: tailwind have lost 80% revenue and laid of 75% of the staff probably mostly due to AI
Lets hope it het sorted out good for @tailwindcss.com it's a phenomenal good product
TLDR: tailwind have lost 80% revenue and laid of 75% of the staff probably mostly due to AI
January 7, 2026 at 10:37 PM
://
Lets hope it het sorted out good for @tailwindcss.com it's a phenomenal good product
TLDR: tailwind have lost 80% revenue and laid of 75% of the staff probably mostly due to AI
Lets hope it het sorted out good for @tailwindcss.com it's a phenomenal good product
TLDR: tailwind have lost 80% revenue and laid of 75% of the staff probably mostly due to AI
I asked Opus 4.5 for an example of how to run this model in Rust, and its solution was to clone a crate to my repository and add a bunch of shenanigans to the code—which resulted in 203 lines of code (even without the cloned crate). My solution was only 3 lines of code using crates.io/crates/rmbg
crates.io
December 23, 2025 at 8:15 AM
I asked Opus 4.5 for an example of how to run this model in Rust, and its solution was to clone a crate to my repository and add a bunch of shenanigans to the code—which resulted in 203 lines of code (even without the cloned crate). My solution was only 3 lines of code using crates.io/crates/rmbg
The cool thing about programming is that we have a free public library we call the internet where we can search on anything we want to build and there is resources for it.
December 23, 2025 at 8:15 AM
The cool thing about programming is that we have a free public library we call the internet where we can search on anything we want to build and there is resources for it.
The best part is that I’ve been able to make this solution very cost-effective by using Rust to handle the image manipulation and model loading.
December 23, 2025 at 8:15 AM
The best part is that I’ve been able to make this solution very cost-effective by using Rust to handle the image manipulation and model loading.
I’m currently building an image CDN that uses machine learning to remove backgrounds. In the example: blue is the ML-powered version, green was my first attempt using a manual algorithm, and light-blue is the original image.
December 23, 2025 at 8:15 AM
I’m currently building an image CDN that uses machine learning to remove backgrounds. In the example: blue is the ML-powered version, green was my first attempt using a manual algorithm, and light-blue is the original image.
Using AI to code is honestly pretty boring, but shipping products built with AI and ML is a blast.
December 23, 2025 at 8:15 AM
Using AI to code is honestly pretty boring, but shipping products built with AI and ML is a blast.
To build something from the ground up and actually see it grow is amazing—I absolutely fucking love it.
With that said, Merry Christmas and Happy New Year, everyone! See you all in 2026!
With that said, Merry Christmas and Happy New Year, everyone! See you all in 2026!
December 19, 2025 at 9:03 AM
To build something from the ground up and actually see it grow is amazing—I absolutely fucking love it.
With that said, Merry Christmas and Happy New Year, everyone! See you all in 2026!
With that said, Merry Christmas and Happy New Year, everyone! See you all in 2026!
Software engineering, programming, and everything in between is one of my biggest passions (though dunking during basketball games is definitely in the top three as well). I spend tons of time every day coding, innovating, shipping solutions, reading, testing things out, and crashing systems.
December 19, 2025 at 9:03 AM
Software engineering, programming, and everything in between is one of my biggest passions (though dunking during basketball games is definitely in the top three as well). I spend tons of time every day coding, innovating, shipping solutions, reading, testing things out, and crashing systems.
2025 is drawing to a close, and I am looking forward to 2026 like hell. It’s finally time to talk about the project I’ve been working on for the last year and a half. 👀
December 19, 2025 at 9:03 AM
2025 is drawing to a close, and I am looking forward to 2026 like hell. It’s finally time to talk about the project I’ve been working on for the last year and a half. 👀
The sad part is that we have made AI "mandatory"/people depend on it so when the prices go up will companies be forced to keep.
Even if the prices is 10x what it's now will some people save money by using AI :D
Even if the prices is 10x what it's now will some people save money by using AI :D
December 19, 2025 at 8:07 AM
The sad part is that we have made AI "mandatory"/people depend on it so when the prices go up will companies be forced to keep.
Even if the prices is 10x what it's now will some people save money by using AI :D
Even if the prices is 10x what it's now will some people save money by using AI :D
Next year, the biggest challenge with all-things-LLM probably won't be building a good solution, but rather maintaining our product's pricing when LLM vendors inevitably hike their rates.
Almost every service offering LLMs via API is still operating on borrowed money.
Almost every service offering LLMs via API is still operating on borrowed money.
December 18, 2025 at 12:39 PM
Next year, the biggest challenge with all-things-LLM probably won't be building a good solution, but rather maintaining our product's pricing when LLM vendors inevitably hike their rates.
Almost every service offering LLMs via API is still operating on borrowed money.
Almost every service offering LLMs via API is still operating on borrowed money.
Go 1.26 will add goroutine leak detection and can already be tested by using expermintal flag goroutineleakprofile
GOEXPERIMENT=goroutineleakprofile go run ./cmd/main
GOEXPERIMENT=goroutineleakprofile go run ./cmd/main
December 17, 2025 at 1:10 PM
Go 1.26 will add goroutine leak detection and can already be tested by using expermintal flag goroutineleakprofile
GOEXPERIMENT=goroutineleakprofile go run ./cmd/main
GOEXPERIMENT=goroutineleakprofile go run ./cmd/main
It's amazing how beautiful dashboards you can build with elixir and phoenix
December 16, 2025 at 1:17 PM
It's amazing how beautiful dashboards you can build with elixir and phoenix
1 thing about "all-thing-sub-agents" where we distribute tons of agents to help us write code is that we think productivity is the same as burning tokens on codegen and at the same time expect AI/LLMs to be the glue to help us maintain this AI codebase we have created with tons of bloated code
December 15, 2025 at 11:42 AM
1 thing about "all-thing-sub-agents" where we distribute tons of agents to help us write code is that we think productivity is the same as burning tokens on codegen and at the same time expect AI/LLMs to be the glue to help us maintain this AI codebase we have created with tons of bloated code
Reposted by Emil Privér
An LLM can't replace good craftsmanship.
December 15, 2025 at 8:56 AM
An LLM can't replace good craftsmanship.