ArqLM update: tools prompts are working, evals regression are coming online, and the inspector is finally getting honest. Screens: Tools registry (file/http/shell toggles), Regression baselines, Providers/MCP settings, Evals dataset import, Prompt editor, and Playground variants. Less guessing. More control.

Langfuse, but it runs on your laptop and never phones home?

*Playstation e Xbox tomando as piores decisões possiveis* a steam:

I decided to cancel the ArqLM Kickstarter. The campaign ended up getting flooded with bot/scam pledges that would donate, refund, and email just to disappear, which made it hard to tell what support was actually real. ArqLM itself is already built and working. This isn’t just an idea or early prototype. What it needs now isn’t fake Kickstarter momentum, it’s real builders using it, breaking it, and giving feedback so it can keep getting better. If you’re someone who actually builds with LLMs and wants to try something different, I’d love to hear from you.

Two things. If this chick had worked for me when I commanded in the Air Force, she’d be in my office at 6 AM, full service dress, and signing her discharge paperwork. Secondly, @petehegseth should issue a directive that social media posts, IN UNIFORM, will result in dishonorable discharge and potential time in Leavenworth. Nip this shit!

I only post when it’s real. Side projects count

the type of video your coworker shows you and asks you if its real

Finalizing a few last things before I start shipping ArqLM out to my early supporters!! Trying to get it to that “stable enough that I won’t have to touch it for a while” point. I'd hate for it to break on day 1. Also trying not to procrastinate bc burnout is real and I’m not tryna crash out 😭

Control isn’t about distrust. It’s about being able to understand and improve the system you’re building.

Been kinda quiet on here bc I’ve been deep in ArqLM 😭 Just submitted the Kickstarter for review. Local-first, inspectable LLM workbench for devs. No magic. No guessing. kickstarter.com/projects/arq…

ArqLM update: tools prompts are working, evals regression are coming online, and the inspector is finally getting honest. Screens: Tools registry (file/http/shell toggles), Regression baselines, Providers/MCP settings, Evals dataset import, Prompt editor, and Playground variants. Less guessing. More control.

GET OUT OF YOUR FUCKING BED OFF YOUR PHONE AND GET ON YOUR COMPUTER

I random queued and got some young kids on my team that said they've never won a game of Warzone. I promised them I wouldn't get off until we secured a W. Hearing how appreciative they were at the end was super heartwarming 🥹. Be nice to the kids man

Suddenly, a savior appears!

Still here. Just on a little break to remind myself I’m building these apps because I want to get good at this, not just talk about it.

Quick clarification on ArqLM. OpenAl is already there. The code knows it. The Ul is catching up. ArqLM is STILL local-first, but providers like OpenAl are supported when you want them ;)