B (@beckyhighness) header
B (@beckyhighness) profile picture

Founder of Your Highness 👑

Los Angeles, CA
Joined November 2021
Photo by B (@beckyhighness)Photo by B (@beckyhighness)Photo by B (@beckyhighness)Photo by B (@beckyhighness)Photo by B (@beckyhighness)Photo by B (@beckyhighness)Photo by B (@beckyhighness)Photo by B (@beckyhighness)Photo by B (@beckyhighness)Photo by B (@beckyhighness)
A lot of the hard work was already done for me by incredibly talented individuals! Please, check out the credits.md that is included. Also, look out for @seanhighness to release even better versions. Im just a simple quality tweaker with my specific strategy.
2
4
467
12gb Cards Stand up!! L0xRE / L0xRE 27b Low Three.js 3d game in one shot on a 3060 Hermes agent / Qwen3.8-27b 8.6gb / Dflash / Vision All running on GPU, Low thinking with 2048k cap on thinking tokens. Stability and Speed improvements paying off
1
4
228
A tale as old as time, now he’ll lose everything and be stuck alone to wither away in his nothingness 🥰
Apparently Lori Huang pawned her wedding ring in 1995 to help keep $NVDA afloat as they were tight on cash She is also the one that picked out the signature leather Jacket for Jensen back in 2013 They’ve been together since they were 17
1
1
2
450
None of my clothes fit anymore.
1
47
Replying to @littlecaesars
@littlecaesars you’re breadsticks are full of these paintbrush bristles that I just purchased and fed to my children.
2
95
@littlecaesars my son was coughing and gagging, and I had to pull this out of his throat!!!!
2
53
Even the cashier said the employees have seen this happen before internally and have mentioned it to higher ups. But they kept the brushes! @littlecaesars
20
L0xRE-27b-Low / L0xRE Runtime 3060 12gb - 65 t/s code 700 t/s prefill Now Uncensored / Abliterated Targets 12gb and 16gb single cards 128k context on 12gb and 256k on 16gb Vision on GPU, Dflash2, Optimized Kernels, Agentic Beast, Windows and Linux / WSL , Uncensored Mode, Works with Hermes, DeepSeek Harness and all other Agents.
1
1
1
6
727
IS THIS OKAY? WTF CODEX So Codex @OpenAI continues using the model for a "TURN" when you switch models, just cost me 4% of my weekly limit as it ran for an hour without actually changing. 166 request sent for an hour without switching ? You're right, and the console wasn't lying about applying the setting — it just didn't take effect. Switching models during an active turn does not change the model that turn is using. Here's the sequence from the rollout file: - Turn 01a17014 started at 15:42:14 PDT and ran until 17:43:55 PDT. - You set the model to openrouter/deepseek-deepseek-v4.1-flash-1017 at 16:33:04 PDT, and Codex recorded it as thread_settings_applied. - The turn context written after that, at 16:40:03 PDT, still says "model": "openai/gpt-6.1-sol". The proxy ledger confirms it. Of the 166 requests sent between 16:33 and 17:43, all 166 were openai / gpt-6.1-sol, with requestedModel: openai/gpt-6.1-sol. Not one went to OpenRouter. That window consumed about 30.5M input tokens (30.07M of them cached) and 116K output tokens. That matches your quota memory exactly: | Time (PDT) | Weekly used | Remaining | |---|---:|---:| | 15:42 | 91% | 9% | | 17:30 | 95% | 5% | The setting is applied at the next turn boundary. Comparable switches in that same thread took effect immediately because they landed between turns — for example on Oct 23 at 16:57:39 the switch preceded a turn start three seconds later, and the next request went to DeepSeek. One correction to your read: the context did compact. At 16:40:03 PDT a compacted event replaced the history, and per-request tokens fell from about 592K to 31K. It then climbed back to 266K over the following hour. So compaction worked; the model didn't change. Your switch is now persisted in the thread's settings, so the next turn in that thread should use DeepSeek. I'd still make it explicit rather than trust the inherited state: codex resume 01a163af-ddf4-7940-9c16-236d0d6baa7c -p deepseek-v41 For future long threads, the practical rule is to make the switch while no turn is running, or start a fresh thread on the DeepSeek profile. If you want, I can write a short check that compares the thread's last turn_context model against your intended profile and warns you before a turn starts on GPT.
5
1
2
506
Best wife ever or.. ? @seanhighness
1
1
45
Pagoda + Hyperframes + Strata / Qwen flash next. Little build video
1
8
461
Our @yourhighnessLA Puffcon 2026 recap, another fantastic event in the books! @seanhighness 🔥
1
1
2
234
Puffcon 2026! 🔥🔥
130
Puffcon 2026! We’re fucking coming!
1
2
119
We're giving away a CONTROL Resonant custom wrapped NVIDIA GeForce RTX 5080 to celebrate its official release with #RTXON. To enter: 🟢 Share this post 🟢 Comment #RTXON T&Cs: nvda.ws/4z24Gcq
37,009
30,519
2,813
33,133
3,115,632
I am introducing the 32 gpu motherboard for Local AI
Introducing an even better motherboard design that quadruples the length of the board so you can natively fit 16 gpus. Longer-ATX
30
12
8
122
13,693
If you're worried about the level of law enforcement retirees interacting with the Cannabis industry just buy weed from the people that have been here forever not interacting with them. 215 is never coming back so the best thing we can do is support the real ones. – at Berkeley, CA
1
4
1
40
1,259
Working on the 0xrca Escha Preview Targeting 12/16 gb cards with a compact high quality version of Qwen3.8-27b that runs fast and comes pre configured.
🤖 Made with AI
1
4
183
I’m 43, Solo builder from Los Angeles Looking to connect with AI enthusiasts and other builders on X. Building with and testing local AI 24/7
I'm 38. Solo founder from Germany. Looking to connect with more builders & indie hackers!
74
1
77
3,223
what will it take to get to 200 followers / friends today ?
86%MEME
14%Replys
7 votes • Final results
2
1
1
180
Celebrating you from the middle of the ocean @seanhighness HAPPY BIRTHDAY!!!! 🖤🚢🥳🫍
1
1
297