antupis

Also tenure Glass would have been just different than blue collar Glass we had.

But usually, in spaces where you need speed Python is just an orchestrator or glue between pipelines, and actually, calculations are done by db or some c/c++/fortran library.

Search is more about data, LLMs are somewhere between those two.

Also, look for those goldrush phases where there is lots of money thrown around eg now AI,

When you have hundreds ci/cd pipelines written on GitLab and people are using ci/cd pipelines more like CRON migration becomes extremely hard.

Well yeah but the problem was that you still needed that runtime, WASM should solve this.

Cost is a plus but at least what I see is that getting good response time is even bigger. Something like OpenAI Azure instances are inconsistent and it is far too normal to get a 40sec lag with responses with gpt4-o.

Silo mainly does consulting and those models were kinda done on side. But great for founders and truly weird acquisition for AMD.

I would say RAG + llama3 is already in that good enough level. Claude and gpt4-o are best but when you need steady performance or a cheap interface on a scale is it very hard to compete against llama3.

It depends on where, like here in Finland those houses that do not have regular or ground thermal heat pumps start to be weird ones.

I would say that being "ok" at managing or planning is easy but being very good at both those skills is extremely hard, so hard that you cannot achieve it with effort alone.

Scrum is good in those cases where it was created so zero trust environment (consulting, horrible PM, etc) and when you have multiple stakeholders. Usually prefer kanban but sometimes you need a pretty rigid process.

> The kind you'd be willing to put on your resume without feeling bad.

Usually when you are doing "just ship it" often enough you achieve this, especially at larger firms too often I see situations where engineers are doing their 3th refactoring without any customer feedback.

Datadog popularity is due the fact that you can throw logs in it and it just works and the dashboard is quite nice same cannot be said about competitors.

That is wrong move personally would start from localllm/llama folks who crave more memory and build up from there.

OpenAI dual structure also makes things super complicated and adds gas to the fire. Like Altman doesn't own anything but has some weird VC fund, and all fuckery around employee compensation.

Well generally yes although there are a couple of exceptions like IoT and GIS stuff where is very common to see 10TB+ datasets.

And The Lord of Rings was first published as separated books.

> Also, why on Earth would OpenAI launch a dramatically better model as long as their competitors don't force them to? The smart solution for OpenAI would be to almost let their competitors catch up to GPT-4 before launching GPT-5 and no competitor is truly there yet.

Is that how Silicon valley has worked like last 20+ years you just deploy fast get customer feedback and then fix stuff based on that feedback. OpenAI holding progress kinda goes against ethos of SV.

Yup sometimes it feels like AWS and Azure are like this.

If you need Google to code in Python, have you really learned Python?

I think 10-15 years is a good guess because the issue is about regulation and infra and these move much slower.

Because plastic from McDonald's is not that big a problem. Fishing gear https://www.theguardian.com/environment/2022/nov/07/invisibl... and carbon fibers are much bigger https://www.sciencedaily.com/releases/2023/07/230703133034.h... plastic polluters.

The logo is very cool I hope someone borrows this.

It would be same playbook that NVIDIA did CUDA where was market 2010 when it was research labs and hobbyists doing vector calculations.

Money, talent, and institutions still cluster around the US and EU where the best universities are so this would hurt seriously AI development.

I think that to get truly working agents we need huge working context windows like 1M - 100M tokens everything before that will be kinda hackish.

It could have a use case in the running LLM interfaces so Apple could probably start scoping the possibility of going after NVIDIA on the server side.

Well Databricks is a big company with real cash flow, and Mistral is a startup so there is a kinda big difference here.

Near Artic Circle there are lots of mosquitoes and moose flies.