frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

The Missed Opportunity for Interactive Book Content

https://zoia.org/posts/the-missing-opportunity-for-interactive-book-content/
1•speckx•1m ago•0 comments

Meta's Muse Tops Poople Bench

https://alexjacobs08.github.io/poople-bench/
1•tacoooooooo•1m ago•0 comments

Why Kids Can't Read Anymore [video]

https://www.youtube.com/watch?v=tj7ckn8WhEM
1•thm•1m ago•0 comments

Show HN: Gibson ADK and Security Runtime

https://www.zeroroot.ai
1•zerodayai•2m ago•0 comments

Perplexity and Nvidia to launch a local AI agent with zero token costs

https://venturebeat.com/infrastructure/perplexity-partners-with-nvidia-to-launch-portable-compute...
1•thm•3m ago•0 comments

Tobi Lutke thinking about banning Claude Code at Shopify

https://twitter.com/tobi/status/2092259436538495186
1•dhruv3006•5m ago•0 comments

Show HN: Bay – open-source Render for AI-generated Code

https://github.com/thebaycloud/bay
1•arsenkk•6m ago•0 comments

Apple Debuts M5 Ultra as Most Powerful Chip Ever

https://www.macrumors.com/2026/08/25/apple-debuts-m5-ultra/
1•thm•7m ago•0 comments

Perplexity Launches Portable Computer Agent Optimized for Nvidia DGX Spark

https://blogs.nvidia.com/blog/local-ai-open-source-models-agents-nemotron/
1•pretext•7m ago•0 comments

When Fonts Fall

https://www.figma.com/blog/when-fonts-fall/
2•microflash•7m ago•0 comments

The Leadership and Architecture Behind Building a Telematics Product

https://vikmodi.medium.com/0-1-on-usage-based-insurance-the-leadership-and-architecture-behind-bu...
1•mooreds•7m ago•0 comments

Show HN: Mindspark

https://prasadpatil25.github.io/MindSpark/
1•prasadpatil25•8m ago•0 comments

Open Source vs. Commercial Auth Providers

https://fusionauth.io/articles/identity-basics/open-source-vs-commercial
1•mooreds•8m ago•0 comments

Australia's music industry bans AI songs from charts

https://apnews.com/article/australia-ai-generated-music-charts-ban-aria-9bfb0c91166ae4405a6df1a3c...
2•MC995•9m ago•0 comments

Warnock: Harnessing GPU Geometry Amplification for Vector Graphics

https://dl.acm.org/doi/pdf/10.1145/3820012
2•coffeeaddict1•10m ago•1 comments

What if AI coding agents had an engineering constitution?

https://github.com/NAEOS-foundation/naeos
1•bayu911•11m ago•0 comments

Argo Events – The Event-Driven Workflow Automation Framework

https://argoproj.github.io/argo-events/
1•mooreds•11m ago•0 comments

Unitree's 45% post-IPO plunge raises concerns over China's robotics bubble

https://seekingalpha.com/news/4636470-unitrees-45-post-ipo-plunge-raises-concerns-over-chinas-rob...
1•ilamont•13m ago•0 comments

How much of HN is AI?

https://blog.coredump.cx/p/how-much-of-hn-is-ai
45•surprisetalk•13m ago•9 comments

Digital Darwinism and the Invisible World of Machine Evolution

https://bryantmcgill.substack.com/p/digital-darwinism-and-the-invisible
1•Anon84•14m ago•0 comments

How to Report a Bug So It Gets Fixed

https://blog.tymscar.com/posts/howtoreportabug/
1•cyndunlop•15m ago•0 comments

CISA slaps its tightest three-day patching deadline on perfect-10 Oracle flaw

https://www.theregister.com/security/2026/08/25/cisa-slaps-its-tightest-three-day-patching-deadli...
2•gpi•15m ago•0 comments

Apple's new desktop computers are designed specifically for local AI development

https://arstechnica.com/apple/2026/08/with-new-mac-studio-and-mac-mini-apple-leans-hard-into-loca...
2•arunbahl•15m ago•0 comments

Show HN: Cinderella – Find out how much value your GH repos have- or buy one

https://cinderella.fyi/worth
1•toddy_flogsit•16m ago•0 comments

Show HN: GPU Accelerated PDAL

https://twitter.com/ZyMazza/status/2092260677301817482
1•zymazza123•17m ago•0 comments

Are public schools unconstitutional? A Nebraska court will decide

https://www.washingtonpost.com/opinions/2026/08/25/nebraska-lawsuit-claims-public-schools-are-unc...
2•mostertoaster•17m ago•2 comments

Super-Intelligence or Superstition?

https://arxiv.org/abs/2408.06602
1•dilawar•18m ago•0 comments

Show HN: See if bloated 'skills' are degrading your coding agent's performance

https://seoagent.com/skill-grader
1•aleclindz•21m ago•0 comments

What Production Incidents Taught Our Team About Authentication

https://www.howdy.com/blog/what-production-incidents-taught-us-about-authentication
1•claudioscheerjr•21m ago•0 comments

Time complexity of operations on Python's built-in types

https://docs.python.org/3.16/library/time-complexity.html
1•theanonymousone•22m ago•0 comments
Open in hackernews

Show HN: A daily cartoon contest between AI models, judged blind

https://comic-cron.com/
1•hahaguffaw•50m ago
Comic-cron.com is a daily contest where AI models draw a single panel cartoon about a recent news story and people vote on which are funny.

I built it mainly because I love cartoons and news, but this project isn’t trying to pretend it’s human art, I wanted to see how capable the models were at this task of making funny comics. Sometimes they get it, and sometimes they don’t and it has been interesting to see the progress in just the few months I have been publishing the cartoons. Model version over version improvement within labs is quite distinct in the output, especially in the SVGs, which suggests their spatial reasoning may be improving faster than their wordsmithing. My briefs and methods have also improved over the weeks so this isn’t something I can purely attribute to the models.

How it works is that a real story is selected, reduced to a shorter brief with any real names removed, and then provided to eight different models to draw a comic and provide a caption. The four of the eight which I think are best are then published where people can vote blind on their favourite (models are revealed after the vote). The human jobs are choosing and framing the story, and then picking the four to publish. Story sourcing is a shell script that pulls a dozen RSS feeds into one file each morning. I tried an agent first and it spent most of its time getting 403'd by publishers; the feeds were never blocked, only the agent fetching them. I initially only included four models to draw the comics (the most obvious labs), but found this wasn’t optimal: sometimes models failed, and generally I felt it was harder to ensure quality and variety with a smaller pool (often models converge and produce jokes or drawings that feel too similar). By producing more panels I could more reliably get a better final four.

Each model is writing an SVG, and they never see the results. This can be interesting in what the models believe they are depicting vs. what actually shows up in the comics, and can be quite funny. Occasionally I will have an image model render the scene as a more polished illustration as a way of rescuing a joke I like that the SVG doesn’t land, or just to show the gap.

A relatively common failure I see is when reasoning models think so deeply about the task and then can’t supply the output because they have spent the token budget before closing the tag on the SVG. Other times the SVGs are just incomplete or don’t populate correctly. I have also had models balk on select sensitive topics (two declined a story that touched on biological pathogens) but complete readily on others (like climate disaster), suggesting there are topic specific nuances for refusal behaviour.

The inference cost has also been quite interesting to track and I often see two of the frontier models accounting for a majority of the cost in a run of eight models. There is quite a variation in the SVG quality, but the captions are much closer and one of the cheapest models is quite competitive.

This is entertainment first and not a rigorous benchmark. But I do hope it gets a few laughs while also being a weird preference poll on something nobody optimises for, that may give a view into how AI models can perform for humour. I’m interested in what people select as their favourites and to see how the models continue to advance.

So far I’ve published daily for over two months and every day’s battle is stored in the archive for viewing with winners and losers. It’s free with no signup and I’m curious on how the format holds up, do the comics land for people? As a news and comic junkie it has been fun to build regardless!