snissn

A) spec driven development

B) opinionated skills that use GitHub tickets, merge gates and execution of ticket graphs

I'm just curious - is duckdb too slow for people? This benchmark from clickhouse shows it being fairly slow compared to some options: https://jsonbench.com/

I’m surprised! I guess I’m being naive but I would imagine you could pass an image to an image model without synthid and have it reconstruct the image in a net new way without the markers. I guess I’m wrong? That’s cool if the watermarks are so deeply ingrained that they persist

I've been thinking about this question for a while. It's confusing at first blush because it's an append-only database and it has a WAL — and it feels like a WAL is already an append-only database, so what's even happening?

Looking back at the project now, I think the value comes from querying it, and especially from automatic aggregations.

    [rollups]
    enabled = true
    checkpoint_file = "rollup.checkpoints.log"
    default_grace = "5m"

    [[rollups.jobs]]
    id = "outside_temp_1h"
    source_metric = "temp.out_dry"
    interval = "1h"
    aggregates = ["min", "max", "sum", "avg", "count"]
    destination_db = "sensors_rollup_1h"
    destination_metric_prefix = "temp.out_dry"

There's also some query support, but that may be closer to something you can sort of do if you just have a log file.

I think the aggregations are the most direct value proposition. OP/author: worth making this pitch "above the fold" in the README, imho.

Also, I've done a lot of analytics work, and a fun feature to add that I've built in the past is an approximate median. I might open a PR and remind myself how to build that. Cheers!

What do you mean by decode is parallel?

it's been frustrating how bad it is at UI. I'm starting to test out using their image2 for UI and then handing it to codex to build out the images into code and I'm impressed and relieved so far

This is weird political propaganda. The first post misrepresented annual costs of housing.

Ok I see so they have six studies that they found to indicate effective treatment of anxiety with cannibanoids but their meta study only found six trials with 50 people so the significance / n count is not high enough. So the study is a null result ie they don’t have enough data not they’ve disproved it. In fact their data supports the hypothesis. Then the coverage on science daily misrepresented this. Here's citaiton 40: https://www.nature.com/articles/npp20116 "Cannabidiol Reduces the Anxiety Induced by Simulated Public Speaking in Treatment-Naïve Social Phobia Patients" Pretreatment with CBD significantly reduced anxiety, cognitive impairment and discomfort in their speech performance, and significantly decreased alert in their anticipatory speech. The placebo group presented higher anxiety, cognitive impairment, discomfort, and alert levels when compared with the control group as assessed with the VAMS.

I’m digging through the paper. It’s a meta analysis paper.. they have six studies they analyze on anxiety treatment and they (footnotes 40-46) have positive results in terms of cbd showing a significant effect on anxiety reduction. I’m not invested in figuring out exactly how this study came to represent what it claims but it seems like a joke

It's not clear that graph-bench in "Tested with the LDBC Social Network Benchmark via graph-bench" is a benchmark that you made. It seems more robust and reliable than "we built a db and a benchmark tool, and our benchmark tool says we're the best". Just a thing to be careful about. You should just state that it's your tool and you welcome feedback to help make it so that other projects being compared are compared in their best light. Something like that might help, I don't know though it's a hard problem.

they should run their test against a control baseline such as an open source hosted model to see the overall drift in their test

https://gist.github.com/snissn/4f06cae8fb4f4ac43ffdb104db192...

I’ve found throw the problem at 3 o3 pros and have another one evaluate and synthesize works really well

I'm impressed with how easy it is to recreate this effect using AI tools. I made this on my iPhone in the photos app by circling Garfield a few times: https://i.imgur.com/YVOZlj8.jpeg based on https://www.gocomics.com/garfield/2025/04/10

to also share a reference - here's an on disk hashmap that uses mmap that I made in golang: https://github.com/snissn/gomap

Human + ai writing tests >> human writing tests

37 million Kelvin

this is worth a read too https://en.wikipedia.org/wiki/The_Noid

Here's a demo of my project shape faces https://shapefaces.com/

new line seperated json "JSONL"

gooey?

do you use the hash of the items you're storing as the key/filename?

anybody have any suggestions for learning openings?

I gave bun a test with react apps and found that it installs node_modules much much faster, but starting up a react dev server took the same time

I would like to get iMessage on my Linux desktop. It’s a step down migrating from my MacBook that I can’t see my texts from my iPhone on my desktop

Cool write up! I had to remind myself these things - nitrogen has an atomic number of 7 (and 7 protons) - Nitrogen-9 means nitrogen with an atomic mass of 9

Therefore N-9 means a nucleus with 7 protons and 2 neutrons!

I’d personally love to get help finding podcasts to listen to. Given a user listening to xyz podcasts show that abc are similar. Or just for each podcast show similar ones with some kind of general explore feature

I like your project. Glad to see alternatives to this https://docusaurus-openapi.netlify.app/

i don't know why people downvote, but writing highly performant gpu code across multiple languages is still in the realm of only a few people with a lot of the right experience can do well, and while ai can help assist those people it's not a problem that can be fully solved by an ai at this moment, maybe a few years with a large feedback loop of iterating, testing, benchmarking, repeating. i guess one day but not now

Hey! I don’t understand enough abt llms. Fine tuning seems like something great but I feel locked out of it. I need to prepare data in a question answer format? I have started to play with taking things like text, articles, tweets and converting them to questions but I don’t think I’m doing best practices. Can you help explain how to take different data sources maybe like a list of documentation for an open source project and fine tune using it?