Defended my PhD this April, and now the dissertation's published--a pipeline for integrating volumetric video and hand-drawn 3D animation, PyTorch and ONNX ML tools for Blender and Unity. Video highlights and link to PDF:
https://t.co/86SJllC9Ux
Also, if you're struggling to find work on Opensea to voluntarily pin, I've been using the Wayback machine to find works to locally store on IPFS. Also, once you have some of the Foundation proxy contracts from Etherscan, you can just keep typing numbers into the Token URI and go into each of the off-chain JSONs, so you can pin both the JSONs and the visual or sonic artworks/assets in the metadata. Anyway, I'm working on a video.
@thekitze Fully automated testing: set up a problem for the agent, give it a way to evaluate the solution, let it keep trying until something happens. This being a pattern that subscription services fire their customers for using outside the metered API makes it intriguingly Streisand-y.
@TheAhmadOsman@1times1time@lmstudiodevs The best local GPUs I have right now are still occasionally needed for VR dev...Llama.cpp backend is running fine for me on Win10 (it's portable, straightforward to set up once and copy to the next box). Are there pitfalls to watch out for though?
@gnukeith Yup, if a potato can run the aarch64 architecture (all RPis 3 and newer can) its general Linux software support these days should imo be good as Intel. For aftermarket Linux on a Windows laptop though, driver support is more likely to be the issue...unstable wifi is a common one
@danzeeeeman Hartmut Bitomsky did a documentary about car manufacturing where someone says, we think automation means one robot does the job of one person, but it really means redesigning the car to be built entirely by robots. Learning how to design automatable code has an inherent value imo
@RohanCreates imo the end goals of 3D creation software aren't going anywhere--pure generation, even used creatively, still pulls pre-existing footage from a very large library, which can be exhausted. But now that we can talk to the apps (MCP) I do expect UIs to change beyond recognition...
@michae1becker I think the post is recycling an older bit of news but the fascinating/scary point still stands--dualism is almost certainly wrong, "mind uploading" really means body emulation
@Justinawblair@lacherbauer@GreeneMan6 iirc Wrath of Khan was a lucky fluke; director Nicholas Meyer was hired without a script and pitched the producers on continuing Khan's story. Roddenberry apparently hated the idea but was overruled. https://t.co/VLRQ1oTtAV
@lacherbauer@Justinawblair@GreeneMan6 imo the irl reason only some Trek series engage with the "dark" canon material--Eugenics Wars, Mirror Universe--is both were created (or at least strongly influenced by) producer Gene L. Coon, who quit during TOS S02 and then died unexpectedly at 49.
@michae1becker @brunomtorrao@henningsanden imo the distinction can get a bit Ship-of-Theseus-y because a common Unreal development pattern is writing new Blueprint nodes in C++
Early Christmas present for the Open Brush community. v2.13 has just been released with tons of amazing new features. Powerful plugin system, .vox and @openblocksapp import, publish to @VIVERSEofficial and lots more. Free and open source forever!
Working on adding MLX/ MLX-LLM / vLLM to Fabric - you can run local LLM models in nodes alongside metal shaders, geometry, compute, realtime video processing, segmentations and key point analysis and make weird shit.
#mlx#llm#vllm
cc @awnihannun
@andybak@kanomalie Compromise: being a good teacher and being fun to hang out with both depend on low-level emotional intelligence, but they're separate skillsets