HN user

js4ever

1,564 karma

I'm working on fully managed open-source solutions for a catalog of 400+ softwares. Also covering deployment and hosting of your own code with CICD. We do revenue sharing with FOSS authors. https://elest.io

Also working on AI Employees: https://Geta.Team

Posts31
Comments563
View on HN
blog.geta.team 5mo ago

Why AI agents need a probation period

js4ever
4pts1
news.ycombinator.com 7mo ago

LLM Benchmark: Frontier models now statistically indistinguishable

js4ever
7pts4
getateam.org 8mo ago

MCP was the wrong abstraction for AI agents

js4ever
4pts1
www.reddit.com 1y ago

My open source project was stolen and relicensed by a YC company

js4ever
45pts4
blog.glitch.com 1y ago

We’ll be ending web hosting for your apps on Glitch

js4ever
213pts139
news.ycombinator.com 2y ago

Ask HN: ISO/SOC2 Certifications

js4ever
5pts8
news.ycombinator.com 3y ago

K2000 in 2022

js4ever
1pts0
old.reddit.com 4y ago

OpenAI Transgender and Diversity bias

js4ever
5pts0
twitter.com 4y ago

ClosedAI

js4ever
3pts0
github.com 4y ago

ImmuDB v1.3.0 – fastest immutable database in Go

js4ever
2pts1
status.figma.com 4y ago

Figma Is Down

js4ever
13pts4
www.protocol.com 4y ago

AI Drones with Taser

js4ever
1pts0
news.ycombinator.com 4y ago

Ask HN: What do you think about Quay.io?

js4ever
4pts4
blog.elest.io 4y ago

Gitlab SaaS Free tier is now limited to 5 users. What can you do?

js4ever
1pts1
news.ycombinator.com 4y ago

Ask HN: Netlify service quality degraded recently?

js4ever
11pts10
news.ycombinator.com 4y ago

Ask HN: How to provide a free trial without being abused?

js4ever
35pts70
news.ycombinator.com 4y ago

Show HN: Elestio – Managed platform for over 150 open-source software stacks

js4ever
108pts51
github.com 4y ago

REST API for Nebula network

js4ever
1pts0
developers.google.com 4y ago

Google Deprecates Sign-In Buttons

js4ever
1pts0
news.ycombinator.com 5y ago

Crazy Crypto Idea: Human Coin

js4ever
4pts1
news.ycombinator.com 5y ago

AMD Ryzen 5950x and Node.js workers = 2M RPS

js4ever
12pts8
github.com 5y ago

Show HN: Lightweight real-time monitoring in Node.js with WebSockets

js4ever
2pts0
news.ycombinator.com 6y ago

Cloudflare removed wildcard SSL in all plans except Enterprise

js4ever
1pts4
www.forbes.com 6y ago

Overwhelming Majority Say Time to ‘Decouple’ from China

js4ever
9pts5
www.bloomberg.com 6y ago

EU Report Says China Deflected Blame for Pandemic

js4ever
13pts0
news.ycombinator.com 6y ago

Ask HN: What is the best web documentation generator?

js4ever
2pts0
websummit.com 6y ago

uFTL, a fast node.js webserver, 500K RPS [pdf]

js4ever
5pts0
www.bleepingcomputer.com 6y ago

EV Certs Are Dead

js4ever
3pts0
news.ycombinator.com 7y ago

AWS Lambda: Node.js 6.10 will be EOL in 40 days

js4ever
2pts0
medium.com 7y ago

A Hacker Wiped Our Business Off Facebook

js4ever
4pts0
GLM 5.2 vs. Opus 1 month ago

"GLM-5.2 hit a problem here, because it can't read images. It isn't multimodal. So instead of looking at a screenshot, it fell back on a hacky workaround: it wrote scripts to read the raw pixel data and check whether the colors came out roughly as expected."

A better way would be to use https://github.com/openbmb/MiniCPM-V

I cancelled my whole GCP account a month ago because I was too afraid of getting charged hundreds of thousands overnight like all peoples on Reddit

MCP is dead? 2 months ago

Totally agree, MCP is the WAP equivalent of mobile internet access.

Thanks for the detailed breakdown. The template header cascade problem makes total sense, I underestimated how bad it gets at scale with heavy template usage. The semiconductor fab constraint is interesting too. When test time costs that much per minute, banning pImpl on hot paths is a pretty obvious call, even if it makes compile times painful. Appreciate the real-world context.

Overhead? None Your real concern is: will potential customers run the model by themselves and skipping us?

Answer is no because you will eventually release a subpar model not your sorta model.

Also people don't have infrastructure to run this at scale (100-500 concurrent users) at best they can run it for 1-2 concurrent users.

This could be a good way for peoples to test it then use your infra.

Ah but you do have an online demo, so you might think this is enough, WRONG.