Invictus Link v2.0 is out.
Phone → your PC. Cursor, Grok, or a local model. Your network. Your keys.
Also wired for OpenAI, Claude, Gemini, LM Studio, and any OpenAI-compatible server — I haven’t personally tested every cloud yet, but they should work. If something’s broken, reach out and I’ll fix it ASAP.
I built it because I wanted it. If it’s useful to you too — take it, change it, make it yours.
Text the app a new look or feature, Cursor ships the update to your bridge, install it from Settings. That’s literally how I built most of this.
Tested on Grok (chat + caching), Cursor (agent), and local Llama.
Grok follow-ups on our setup: ~57% cache hit, ~43% less input cost vs no cache.
Clone the repo, open it in Cursor, ask: “scan the workspace and walk me through set up.”
Enjoy!
>be elon
>"we are in the Singularity"
>goes ALL IN: AI hardware & software
>january 2026
>anthropic cuts your claude access
>“safety concerns” they say
>hmmm ok
>feb: grok 4.2 drops (delayed)
>timeline says it’s mid
>merge xAI into SpaceX
>drop grok 4.3
>timeline says it’s over
>you have the biggest AI training cluster on earth
>but sitting at only 11% utilization
>mainstream media dropping hit pieces
>elon: “I will never give up. Never”
>sees cursor sitting in plain sight
>the default coding harness of corporate america
>7 million+ hardcore swe coding all day
>every keystroke a labeled training example
>the entire market prices it as an IDE
>*look closer*
>composer team already turned a kimi base into a real model
>proof of a real RL training, infra and data team
>hopium
>april: cut the cursor deal
>$60B option to buy, $10B just to date first
>they say yes
>give them access to colossus 2
>gb300s, the best AI compute on earth
>tell them it’s time to lock in
>*suddenly phone rings*
>it’s anthropic
>“we need your datacenters”
>we will pay whatever you want
>kek
>lease them all of colossus 1
>$1.25 billion per MONTH through 2029
>add a clause: we have the right to cut you off anytime
>two weeks later google calls
>$920M/month for 110k more
>$26B/yr contracted. more than ALL of spacex 2025 revenue
>meanwhile cursor team proves they can cook
>spacex IPOs
>sign the cursor merger, all stock
>welcome to SpacexAI
>drop grok 4.5: big leap
>haters suddenly change their stance
>“xAI might make a comeback”
>true believers (me) know: WE ARE SO BACK
>drop new product: grok bot
>actual digital co-workers
>chatgpt work and claude co-work on alert
>next day drop grok 4.6
>frontier model
>ties with claude fable on performance
>beats gpt sol on frontier code 1.1
>cheaper than deepseek in task completion
>timeline shocked :o
>elon reminds them grok 4.7 has finished initial training
>now adding massive spacex company data
>dropping in 3-4 weeks
>“This will be something special”
>this morning: cursor deal officially closes
>largest startup acquisition in human history
>pay the whole $60B in stock
>wall street calls it "the deal of the century"
>they don’t know
>this is just the beginning
>“AI will become 99% the value of SpaceX, and SpaceX’s value will be astronomical”
ITS HAPPENING
@rediscoding Maybe some articles or lectures on AI use, development, stacking etc etc! How to train your own models, free and paid resources for doing so, stuff like that!
These 10 startup accelerators will write you a check this month (up to $1M) and anyone can apply. Deadlines are live.
Bookmark this + send it to every founder still grinding 🧵
Share your favorites in the comments!
Invictus Link v2.0 is out.
Phone → your PC. Cursor, Grok, or a local model. Your network. Your keys.
Also wired for OpenAI, Claude, Gemini, LM Studio, and any OpenAI-compatible server — I haven’t personally tested every cloud yet, but they should work. If something’s broken, reach out and I’ll fix it ASAP.
I built it because I wanted it. If it’s useful to you too — take it, change it, make it yours.
Text the app a new look or feature, Cursor ships the update to your bridge, install it from Settings. That’s literally how I built most of this.
Tested on Grok (chat + caching), Cursor (agent), and local Llama.
Grok follow-ups on our setup: ~57% cache hit, ~43% less input cost vs no cache.
Clone the repo, open it in Cursor, ask: “scan the workspace and walk me through set up.”
Enjoy!