BlockedPath

Alright up to 7 3090's on my machine, most recent through m.2 all at x16... it's SUCH a rats nest lmao. So now I have a machine with at least 2 free pcie slots (the machine I took the 7th 3090 from).. Have to add another here to make it 8, and 2 on my empty node now!

@evi77ain Hooks

I think I've found a cure. And it's simple, really. In my setup, Codex Desktop is using MultiAgent V2. I'm not sure whether that's the default for everyone or something in my config. Sol and Terra are marked as V2-compatible, while Luna, for some reason, is still marked as V1. So `spawn_agent` simply filters it out. If your CLI or Desktop App also uses MultiAgent V2 and Luna is missing from `spawn_agent`, you might be running into the same mismatch. You don't need to switch the whole app back to V1, btw. You can just patch Luna's catalog entry yourself, actually. 1. Copy: `~/.codex/models_cache.json` to: `~/.codex/model-catalogs/desktop-multi-agent.json` 2. In the copied file, find `gpt-5.6-luna` and change: `"multi_agent_version": "v1"` to: `"multi_agent_version": "v2"` 3. Add this at the top level of `~/.codex/config.toml`: `model_catalog_json = 'C:\Users\YOUR_USERNAME\.codex\model-catalogs\desktop-multi-agent.json'` Use the absolute path, obviously. 4. Fully restart Codex Desktop. I've checked the actual child rollout, not just whatever the model claimed in its reply: GPT-5.6 Luna, Max effort, MultiAgent V2, real subagent, non-empty parent thread. It works in my setup. One annoying catch, though: `model_catalog_json` replaces the entire bundled model catalog. It's not an overlay. Which means your copied catalog can become stale after a Codex update. So don't forget you've changed this. After every update, try spawning Luna without the override first. If the mismatch has been fixed, remove the `model_catalog_json` line from `config.toml`, delete the copied JSON and restart Codex. If it hasn't, rebuild the copy from the fresh `models_cache.json`, change only Luna's V1 → V2 flag and restart again.

@loktar00 6 months!

Grok ai sucked me back in with the discount again and somehow I got the $100 for 6 months?! Worth it with 4.5 and the speed, excited for 4.6.

@Daniel_Farinax @WOwenWebb Im really thinking of taking this over codex 100 plan right now.

@WOwenWebb legend

SuperGrok Heavy is back on sale for $99 🔥 I snagged it at this price 3 months ago and it’s been 100% worth every penny.

@mikenevermiss cli api proxy or cc switch

this guy just wired kimi k3 directly into codex. the model is insanely good, but the app isn’t. so instead of using the kimi app, he connected his kimi subscription to codex through oauth. now kimi k3 shows up right inside the model picker alongside sol 5.6. no api key needed. no extra subscription either. just sign in with oauth through kimi code, and it’s ready to use. simple guides in the video and links below 👇

@ChrisIzatt Heck yeah buddy, looks so clean

Genuinely impressed - finding it was surprisingly difficult to actually max-out every sub (so I can get some chill time!) Grok & Cursor here we goooooo!

@pidotdev @herdrdev @mattpocockuk Here is my theme settings gist.github.com/BlockedPath/c8…

For a great workflow setup I personally love using @pidotdev , @herdrdev , and matts skills @mattpocockuk . Simple, clean, and effective.

@xiaoxin880227 @pidotdev @herdrdev @mattpocockuk gist.github.com/BlockedPath/c8…

@BlockedPaths @pidotdev @herdrdev @mattpocockuk This pi theme is amazing, could you share it?

Maybe I was a bit harsh and biased with this, as I am only human. There are edge cases where its appropriate to use it, as well Hermes. I just wish the opensource community wasn't so divided at times. I'm hypocritical. @pidotdev will make you learn how to do things

@wolfie_ ouch my hears lol

OMPbrrrrt v2 - sound on I added a distinct tone for tool calls

I made a little add-on that lets omp search X for you. It pulls actual X posts with links, then gives you a straight answer. No API keys, but it needs your existing xAI (Grok) login (can work on X Premium, dont need supergrok) github.com/bnivanov/omp-x… Install is one line, instructions are in the README. Once it's in, just ask your agent to search X. It might be slop-y so looking forward for fixes if needed. Thanks @BlockedPaths for the inspiration and guidance with the pi plugin @usr_bin_roygbiv @oleksoleksoleks

@bnivanovX Cheers, nice work bro!

@LexnLin Been having better luck using pi and Claude with 5.6

Testing models from other providers in Codex 👀

@jturntdev I also noticed this a few days ago. Went from being nice and steady to gone in a blink of an eye

Cant just be who’s noticing this but GPT 5.6 sol in general drains your quota extremely fast. How is it more efficient? On 20x plan, you now get 1-2 days of usage. If you’re being conservative. We’ve had alot of resets, this is just to stop everyone complaining. When they all go, i see the beginning of the end. They also reset limits 3hrs after we had to use our banked reset, due to expire.

@bradshannon @miniii_codes Nice write up and valid points across the board

@miniii_codes This might help... x.com/bradshannon/st…

We have 500 verified followers now. 🎉 Just one step away from monetization. But honestly, I don't know if I'll ever reach it. Between a full-time job and responsibilities at home, finding enough time to create consistently is tough.

@samsaffron @grok I don’t think so, every harness has it at this point. Just limited, I made this 3 months ago-ish

@BlockedPaths But does @grok bless using its oauth dedicated creds in third party apps?

pi-xai-oauth v1.3.6 is ready 🚀 - Its better than the native xai auth, which doesnt have native tools. ✅ Account-specific Grok model discovery ✅ Browser + device OAuth ✅ Encrypted reasoning replay ✅ Grok-native coding tools ✅ Image editing + usage tracking ✅ Hardened routing, security & entitlements ✅ Pi 0.80.1–0.80.10 support pi.dev/packages/pi-xa…

@maxedapps It’s a good deal, and 4.6 is about to come out. 4.5 is such an amazing model for its speed and power.

Do I know anyone with a supergrok $300 plan? How much usage do you get out of it compared toto openai $200? Obviously a bit hard to tell with all those constant limit resets but would love to get some insights

@Skiipy88 Sweet sweet ram

Paid 900$ usd for these bad boys, 192gb ddr4 ecc 3200. Marketplace hunting still full of good deals, if you’re patient👌

@DJLougen @usr_bin_roygbiv 😂

@BlockedPaths @usr_bin_roygbiv

If you’re using omp, stop and take a second to realize it’s pure slop and a terrible harness wrapped in bandaids.