Very low quality content. Unsurprisingly pangram 100%.
I just write this way lol, hard to be "pangram 100%" when I'm getting criticized for typos in another corner
Recreating Minecraft Is Not a Benchmark
61–70 of 81 posts
Re: Recreating Minecraft Is Not a Benchmark
#62It’s a tiring YouTube formula:
“Model X CHANGES EVERYTHING!” (surprised Oculus Rift face without the headset)
Some one-shot prompts for quick impact.
I get it with the incentive structure — honestly not trying to shit on any creators (hate the game, not the player) but seeing it every day is just getting tiring.
Re: Recreating Minecraft Is Not a Benchmark
#63Earlier quoted context omitted.
Important distinction, but I would debate the "most common tutorials on the internet" claim.
I think you meant “pedantic distinction”. Everyone reading that comment understands that recreating Minecraft is a very common tutorial for game development, not literally one of the most common tutorials on the internet in general.
We are on HN, I thought that goes without saying.
Re: Recreating Minecraft Is Not a Benchmark
#64Earlier quoted context omitted.
> ridiculously polished detail Can you point me to one of these? The only Astra game I tried was the incredibly janky Mario Kart clone from OpenAI where you could fly by spamming jump
I was impressed by this 3D version of Zork. Video: https://bsky.app/profile/emollick.bsky.social/post/3muqcwa4j... Demo: https://zork-underground-empire.netlify.app Code: https://github.com/emollick/zork-underground-empire Prompts in a screenshot here: https://bsky.app/profile/emollick.bsky.social/post/3muqdyefc... - I extracted them as text here: https://gist.github.com/simonw/dab1843ef17edeaeffd18b0acbba2...
that is actually REALLY cool, wow
Re: Recreating Minecraft Is Not a Benchmark
#65I keep seeing Astra make beautiful 3d stuff online, yet when I feed it some old school RuneScape assets (even tried with some very detailed guidelines) and asked it to generate some new plausible assets it failed horribly. I think there’s still something really off with current (frontier) models when it comes to creating “novel” stuff? Even 2004 style graphics.. Or am promoting it wrong?
The type of people that can’t get results are the people that don’t share the process that didn’t work for them.
Re: Recreating Minecraft Is Not a Benchmark
#66Most of my and my peers PortCos run their own eval and benchmark sets, simply because they know what they need best. The reality is, capabilities have largely converged across foundation models over the last 18 months, and much of the value add is coming from the harness layer itself now. This has been the operating assumption for me and my peers, and has largely played out that way. That said, this has always been a…
> The reality is, capabilities have largely converged across foundation models over the last 18 months, and much of the value add is coming from the harness layer itself now.
Can you please elaborate on what are your thoughts on open weights models (GLM 5.3, Kimi K3, deepseek etc.)
and if the value add is coming from the harness layer itself, then thoughts on open source harnesses (there are so many harnesses but to name a few: opencode, pi [omp as well], maki, codex is OSS as well, fx.sh) and you can always combine them with skills (Obra/superpowers, matt pocock skills plus using these skills and others to create some other custom skills tailored to your use case as well)
And what about the combination of both now with this cheap open weights models + open source harnesses and other things to compete over the closed garden ecosystems?
How does that comparison follow in reality
Could you in theory use these methods to save on the massively expensive $$$ token spending on Anthropic/OAI?
(Personal anecdote but I have GLM 5.3 + maki [sometimes omp/opencode but mostly maki] and its good enough for most use cases out there that I have and I dont know of too many use cases outside of say recreation of games for examples maybe that I would prefer complete SOTA models. I would also love to know where you believe that SOTA models absolutely do still make the difference discounting the benefits provided by the harness.)
Re: Recreating Minecraft Is Not a Benchmark
#67Earlier quoted context omitted.
Impressed like that's cool ai is doing it? Or just impressed in general like its high quality?
I think it's a genuinely good implementation of the concept. It looks good and plays reasonably well and captures what it needs to capture from the original. (Update: OK the sword fighting isn't good. As I got deeper into the game my positive first impressions wore off.)
I haven't tried any of these demos, but I'm not surprised they stop impressing once you go deep.
What got be mostly impressed were the demos of Astra doing computer use. At my job I do some RPA and can appreciate how challenging it can be. Yet they make it look extremely easy to operate a tool like Blender at super human speeds.
Re: Recreating Minecraft Is Not a Benchmark
#68Earlier quoted context omitted.
Works on phone but not on my linux desktop.
Mac here. There's nothing all that interesting going on with the JS on that page, so I would suspect a bug with Firefox and/or your desktop environment.
Codex found the root cause:
> The site’s graph incorrectly uses 32 texture bindings, exceeding its WebGPU device’s limit of 16. > Firefox 153 mishandles that failed rendering operation and leaves the compositor waiting
So a firefox + website bug
With an updated firefox, the website works except the mindmap on the top right
> Uncaptured WebGPU error: Too many bindings of type SampledTextures in Stage ShaderStages(FRAGMENT), limit is 16, count was 32. Check the limit `max_sampled_textures_per_shader_stage` passed to `Adapter::request_device`
Re: Recreating Minecraft Is Not a Benchmark
#69But they're not recreating Minecraft. They're recreating one of the most common tutorials on the internet that shows a simple voxel overworld and nothing else. That's not Minecraft.
Only Minecraft is Minecraft. Is that the bar for creating Minecraft?
Re: Recreating Minecraft Is Not a Benchmark
#70>GPT-6-Astra-Max
"Generate a SVG of a PlayStation 4 controller as nicely done as you can"
Time taken: 11m 54s
No image was used
I was expecting 10s not 12m; if it takes 12m to do it, I would rather try to do it myself. And what it means "No image was used", wasn't it trained on bunch of images including probably console controllers images?