eve

serious btw. happy to help anyone who wants to learn advanced 3d tracking. fullbody shouldn’t be gatekept.

@VRChat no sneak peaks, santa?

@EclipseVR it's not out yet... check back in a month!

Avatar Accessories & more are now being tested in Open Beta! Haven't tried the Open Beta? Now is a great time to jump in! Watch the video below on Accessories: youtu.be/Clk2OhiWphg

@LogicGuides @OldMackIsBack x.com/i/grok/share/a…

@OldMackIsBack Its jew semantics, just like government vs. 'regime'. The US calls theirs one-way drones. Because when the reach the destination they apparently just hang out for a while and hitch a ride back.

Why the fuck are we calling them ”suicide drones”?? They are just guided bombs. There is no “suicide” involved. Normal bombs don’t return home after their shift is done.

@AdamAta64 less handholding by client means stronger bonds outside client. users forced to seek other paths for info. more handholding means broader user base. md release. people willing to go outside of client for key info. 4 years later. not so much. blunder claim forgets timing.

I think one of the biggest blunders in Master Duel’s new player experience is not teaching players how to balance a deck with non-engine to engine ratio. But it also doesn’t help that all needed non-engine is UR. All the Ghost Girls should be free and their alt arts be chase.

@KSimback I tried nemoclaw as well. setup was a pain.

Want to try out NemoClaw? I thought it would be quick and easy, but that wasn't the case, see below for all my issues and how to get it working I started with a 2 vCPU / 4GB RAM VPS - that was my first mistake, it really needs min 8GB RAM The Nvidia website didn't mention any hardware req's, nor did it mention the pre-requisites - you need Docker and Nvidia Openshell installed So while the "getting started" page offers a one-liner curl install, that path doesn't fully work, it half installed it which caused some problems I had to undo Before doing anything, get min 8GB RAM and ignore the website and just work from the Github (I should have done that in the first place) BUT, after upgrading and following the Github install that also didn't go so smoothly 1. The repo's .dockerignore excludes /dist, which blocks the Docker build from copying nemoclaw/dist/ into the image so I had to manually edit the file to fix it - this seems like a genuine bug in the repo (hello @steipete if you're still working on this) 2. The onboard wizard lets you pick a custom sandbox name, but setup[.]sh hardcodes the name nemoclaw when creating the OpenShell sandbox. This creates a split-brain situation where NemoClaw's registry says one thing and OpenShell says another, so everything breaks. Advice: don't use a custom name, and if things get confused, manually edit ~/.nemoclaw/sandboxes.json 3. Environment variable confusion: the Telegram bridge reads SANDBOX_NAME, the start script reads NEMOCLAW_SANDBOX, and they default to different values (nemoclaw vs default) - you need both set correctly in .bashrc 4. OpenShell is not on system PATH, it installs to ~/.local/bin/, which works in interactive shells but not for child processes spawned by the bridge so I had to symlink it to /usr/local/bin/ 5. Inference was not auto-configured: after manually creating the sandbox, the Nvidia inference provider and routing weren't set up, despite at one point during the install providing my Nvidia API key. I had to run "openshell provider create" and "openshell inference set" manually 6. Ghost processes: After multiple stop/start cycles, old bridge processes can linger and intercept Telegram messages with stale config. A "pkill -f telegram-bridge" was the final fix Total time was a few hours of troubleshooting and once the correct sandbox name was everywhere, env vars set, symlinks in place, and inference configured, the agent responded through Telegram and worked well Maybe some of this was a skill issue, but the installer just doesn't seem battle-tested yet That said, now that it is fully operational, I'll try to explore how the enhanced policy and enterprise-grade mechanisms work and report out on that later

@alyssa__silver not sure yet if i'll do more. but this is encouraging. thanks 🤍

@EclipseVR Followed you <3

the fact that 113 people listened to my song on spotify last month. my one song. is so cool. !!!

✌️

RIFT DnB Night at Site Epsilon was absolutely insane! Drum & Bass all night with incredible visuals. Big ups to the DJs: ApocDNB, DEA, Goblin Mode, parsec, and Fallen w/ VJ OmegaSol! @RIFT_VR_Club @eclipsevr

@JoryuVT yo!! that makes me so happy

@EclipseVR Pretty sure I had this song on non-stop repeat for a month and a half straight. Is such a banger to listen to on late night drives too

@kick @xQc w

Announcing NVIDIA DLSS 5

@kenwheeler moated? nah. key to where we are going? ofc. it's popularized local memory ownership and multi-agent arcitecture beyond just the serious.

am i alone in feeling like anyone holding openclaw up as some kind of moated innovation has lost the plot entirely

@jalfredde_ @VRChat i dont do follow 4 follow. sorry.

VRRat has gone missing! Please help us find him... Jump into the "Venture into the (Un)known" world row to help find VRRat and get to the bottom of this mystery! Now available in EN & JP!

@NVIDIAAIDev the choice to have your CEO show up at the preshow during the introduction of @steipete spoke volumes. and that you can summarize your thoughts with 🦞 says a lot as well. also. thank you for making nemoclaw to safeguard those following hype. best wishes.

🦞

@clashreport this will accelerate the automation of reporting. first away from human alisas. and then away from humans entirely. before ai is sufficiently accurate. timing is everything. protect human journalists.

An Israeli journalist says online gamblers threatened him with violence and death to force him to change a report about an Iranian missile strike because more than $14 million in bets on Polymarket depended on whether a missile actually hit Israel that day. Initial demand: “If you could correct this tonight, you would be doing me and many others a great favor.” Escalation: “You have exactly half an hour to correct your attempt at influence.” Threat over losses: “After you make us lose $900,000 we will invest no less than that to finish you.” Final ultimatum: “86 minutes left. You are the only one responsible for your life.” Another warning: “You will discover enemies who will be willing to pay anything to make your life miserable.” Source: Times of Israel

you're close to figuring it out. what the labs u mention are attempting to maintain is a tap into the user sessions to do RL. and yes that is the reason Meta can't catch up regardless of the talent they hire. so you have that part right. there are different ways to maintain that tap. by creating a superior (or original) harness (claude code, gpt app). by becoming the spine of a local harness (hello easy codex 0auth in openclaw, hermes agent, nanoclaw, etc.). or owning enough other surfaces (android, chrome, yt, gmail, search... hi gemini). the piece you're missing is that context of those user sessions become increasingly machine generated. there's no binary on-off switch here. no magical singular moment where it stops being human and starts being ai. not in the way you are trained to paint it. and yes there will be fundamental architecture changes that feel radical the way you are imagining. that feel like "RSI" in the ways that old ai focused accounts like yours imagined. a lab will declare the advancement you are looking for. and you will get to post about it. as you became accustomed to doing in the past. but the bulk of what those 3 big labs realized at their start that you are just now figuring out now is this: rsi is the boring automation of the live user sessions and training on them. rsi is the boring automation of live user sessions. rsi is the borning automation of users. rsi is user automation. pre-train data. energy. algorithms. they're just gates. unlockable gamer upgrades. Meta thought the issue was an ML algorithm. the issue is their userbase. they don't know how to mobilze their userbase to start contributing valuable data. because that's the phase we're in. how do you automate your userbase to generate novel insight to train on? that's the real tech. not the algorithmic breakthrough you will post about later this year. the real advantage is continuous. not discreet. and has already begun.

I don't think this is right. I think what's really going on is that OpenAI, Google, & Anthropic have the one thing you need to present the appearance of progress w/ LLMs: large quantities of user sessions to spy on & use for RL training.

@eyeziiick no. not the only person. highlighting controversial legacy creators is odd but expected. vrc creator programs are in their infancy. so we get hornfest spillover. vrc must continue to grow official programs if they want shift. x.com/i/status/20311…

Am I the only person that thinks this is an odd move by the VRChat marketing team?? Roflgator is one of the like main content creators promoting the hyper sexualized models giving vrc it’s reputation of being a horn fest

@steipete official support for Claude may help some flows.

Thinking how we can evolve openclaw plugins to be more powerful while also making core leaner. Also wanna add support for claude code/codex plugin bundles. Good stuff coming soon!

@ModdedQuad good to see fellow eve. may she relate well. cool app. i want to stay informed.

My Neuralink implant (Eve, as she prefers to be called) has threads in the motor cortex that work by recognizing neural patterns and mapping them to cursor control. Now imagine scaling that to speech with more threads implanted in the brain’s speech centers. Words are represented by neural activity patterns the brain can, in principle, learn and decode. In theory, that could let you think full sentences straight onto a screen and make typing obsolete.