jerrygenser

Wouldn't lazy loaded relationships mean that every time you access one of the entities it lazily loads, leading to n+1 fetching behavior?

What's wrong with lambda? If you write a monolithic app that serves your rest API via lambda that pattern can be very powerful. What alternative for such a use case would your recommend instead?

It is 1.0 only because of supporting pydantic 2.0. It's not 1.0 because it's done and you should try to use it. In fact, 1.0 is unstable I don't recommend upgrading if you are running django ninja in production

This concept is featured in The Dark Forest, the second book in the Three Body Problem trilogy. It was something where you could genetically tag someone and then if they woke up 50 years later from a chrono sleep, various systems were preprogrammed to have "accidents" around them, like self driving cars, etc.

Love the expected noise while in ludicrous mode

There are queues like sqs and google pubsub where there is already official clients written for most popular languages. And they also have http API so you can use it in any language environment that can make http requests.

If you don't need SQL (relational data), but maybe have a schema per topic, I've used rocksdb as a cache for latest in tombstones topic. It has high write throughput for rebuilding state when playing forward a stream

Adobe is already incorporated genAI into core product and will start pricing for compute units. This is part of their core product, not a tool being sold to other people to build genAI products

LLM are not somewhat fast when measured against CPU models in spacy that are specialized for a task. They are multiple orders of magnitude slower for e.g. token classification/NER, text classification, etc.

that's not to mention the plumbing that spacy provides that are mostly bindings to C code for tokenization, lemmatization, etc. things like that which are more algorithms problems than machine learning problems.

All (or nearly all) of the top pros were using PED for a number of years.

I have a similar use case and I have had success reading from sqlite files embedded in my repo

I think it's still overkill though for semantic embedding, SBERT is on order of ~250mm parameters while smallest llama at 7b parameters.

Looking for a single custom prompts that can be used across many specific programming questions.

I think your example is more like how to write a prompt for a specific question

I clicked the about link in the top right. I was pleasantly surprised that it was quite a good advertisement for cloudflare workers.

Health insurance, workman's comp, pension liabilities, etc.

The "fully loaded" cost of an employee is a _lot_ more than the cash comp they receive

Sure, I bring it up though because it's discussed in those behavioral psych books.

I'm saying it's good enough for my use, which is watching YouTube videos, using apps, taking pictures, having a reasonable battery life, making phone calls, sending texts.

It might not be the best possible soc etc. But for $349 it serves me as a daily driver for 3 years without shelling out over $1000

There are some types of nudges that have had measurable and durable societal benefits.

The primary one I'm aware of is making 401k a default, people are likely to stay enrolled and increase enrollment.

Funny, my last android was samsung which I had many of the issues you describe above, and in particular it felt very bloated. I've been getting older generation Pixel S and they are very good for the price. usually 1/4 the price of a comparable flagship iOS and I never have any issues.

You are literally using trillion documents? Or are you exaggerating?

Llama cpp speed is dramatically improved by avx instructions. If your CPU has those it would be much faster than not.

And if it doesn't you need to do some workarounds with compiling and it gets a bit harder to run.

There are a bunch of OSS that solve problems when running distributed scale out workloads in "the cloud". For example prometheus is a metric db for storing monitoring data.

Ceph is a storage backend. Most of them tend to be built for or work well with kubernetes.

Just a note I believe this is a reference to the book "what I talk about how when I talk about running" by mirakami

Slightly off topic. I used to play materiamagica a text based mmorpg. My main character had 3k+ hours on it by the time I graduated high school. But I played some more during summers in college.

Anyone else play them?

Refuting the points in (a) and (b) still concedes that it was made private. Care to actually mention _why_ it was made private?

I think using this project https://github.com/ggerganov/llama.cppav

on a CPU machine with AVX instructions would be a better bang for your buck than GPU. Depends on if your use case can tolerate the latency

Probably quantizing or using base weights and this project https://github.com/ggerganov/llama.cpp on a CPU machine with AVX512 instructions.

Not meaning to be controversial, curious - why is it under a16z-infra namespace?

I see it can be used to define and run workflows. But for reproducibility on top of the execution of operations on data, I'm wondering if there's any way to version the underlying data?

Or would you typically use this -in addition- to a tool like dvc? I've used dvc a bit and while it's quite good for data versioning, I find the workflow aspect is clunky.

They are selling wood and metal to make shovels, which is substrate.

There are shovel sellers which are making tools.

Then there are application makers that end users see that are made with shovels.