I just joined X. Time to reintroduce myself
I run an MSP, build indie games, & explore AI, automation, self-hosted tools, game dev, & other tech rabbit holes. Big sci-fi & gaming fan too
Always happy to connect w founders, devs, creators, IT folks, and builders
Come say hey
Tabby Terminal has changed my life - just discovered it and it is sooooo good. Opensource, quick SSH terminal with drag and drop SFTP, multi-tabbed. What tool are you using to manage your nodes and instances?
Here's the full audit for others who have a similar issue - the audit also surfaced and references the third party launchers as well as stores for game assets too.
https://t.co/BRpnoAHo6A
I have had ticket w @EpicGames to fix my launcher for months - can't install any games, can't reliably use UE & my assets. Also been trying to install 5.8, 2 days now, it's greyed out. So put @openai codex on the job. Turns out I have too many assets for launcher api to manage.
Been trying to install @UnrealEngine for the last 2 hours - 5.8 and it does not download. They really need to fix that @EpicGames launcher already! It's such a terrible and bad experience for such an innovative company.
#unrealengine
web ui to my #turboquant forked #llama variant. I added a #huggingface search/download manager to list models based on hardware. Included realtime stats on cpu,ram, gpu vram, along w configurable param for models. adding RPC to balance models across network to other llama nodes.
WOW got 35b model loaded on old rtx 3090 using llama w modified #turboquant. It's insane. It's not even taking huge amount of vram cause it swaps the moe specialist to cpu & ram w 250x cnt. Running on threadripper with 128gb ram and rtx 3090. I can't wait to try this on my 5090.
Seriously what is going on w #Claude - I pay for max plan & it feels so dumb lately compared to Chatgpt. I finished 2 big projects w #codex today but struggled so much w just 1 project w Claude today. Even simple tasks like my scheduled tasks, I had running fine, bombed today.
And it produces an A.I. Remediation Pack for your repo with it's findings so you can send in your preferred agent to clean up the code of any forked repo, etc once you verify or remove any found malicious code. It's a standard markdown report you can feed to your A.I.
So I built gitsafe internally to scan repos before using them locally or in the lab and wow - there are a ton of malicious, gitleak, badly formed execution projects on github - definitely helped me so far. I even built in AI triage support into the audit. #github
I also added terminal into the repo to check out the files right from the dashboard - so you can inspect the repo, run agents etc. The terminal shells into the sandbox and is guard railed for your local machine security.
I started using Xiaomi MiMo V2.5 Pro today as I have been testing different models over the last week and it's surprisingly very capable and fast model for openclaw or Hermes for complex tasks
#xiaomi#ai#llm#openclaw#hermes
Been playing with various AI coding harness this week - goose, pi, kino, cline, opencode so far - I really like goose and kino so far. What do you guys use and why?
#agents#coding#AI#homelab
Made tweaks to gitsafe app - clone, sandbox #github repo, scans repo w tools: Gitleaks, Semgrep OSS, OSV-Scanner, TruffleHog, Trivy - ability to terminal into the sandbox to work or inspect code. Offers AI remediation instructions also to give to AI after the scan is clean.
Built repo scanner for #github - I'm constantly trying out new software, packages - wanted a selfhosted solution where I could add a repo - it creates sandbox, clones repo, scans & looks for issue & offers remediation packs per repo