the real AI race is about building the best infrastructure: electric grid, inference, training farms, power generation. there is no race with building models. models are commodities. infrastructure is the moat.

@Kai how can i take out a flock camera

as long as openai subsidizes 5.6 sol no one will feel compelled to pay for nerfed fable 5. as long as open source keeps catching up, openai will have to keep subsidizing its frontier models. then the big companies run out of cash and open source wins.

apparently door dash gave every employee claude code. the odd thing is, two employees with claude code could create and run door dash themselves

claude now refuses to do anything related to security. tried having it write a script to remove encryption from an old zip file i had lost pass to and it's refusal claimed i was breaking local laws. within minutes GLM5.2 does brute force with gpu acceleration to decrypt. anthropic is on death watch 💀

llms will always code the best on opensource software because the entire project data is readily trained on.

bitcoin is digital 1776

great time for everyone to use an LLM to help them learn nixOS nix-btc and the power of running your own lightning node. level up this month View quoted note →

voltage is basically shutting down for normal people running a node

We're Sunsetting Self-Serve at Voltage: Enterprise is Our Future | Voltage

Voltage is deprecating self-serve to go all-in on enterprise Lightning infrastructure. What it means for existing customers, the key shutdown dates...

exact replica of claude code is achieved with very easy steps. 1. install opencode 2. enable lsp

OpenCode

LSP Servers

OpenCode integrates with your LSP servers.

2. enable exa webfetch

Getting more out of OpenCode

OpenCode feels rough out of the box, but a few setup tweaks make it far more capable for agentic coding

3. use no logging glm5.2 model on deepinfra 4. ask claude to create a open code command that exports its project memories and agents.md for each of your projects into opencode when you run the command 5. launch opencode in a directory of a project and run the command claude made and bring over your agents.md files 6. enjoy being free of anthropic with mythos level coding

opencode + glm 5.2 on a private inference provider (deepinfra doesn't log) has just replaced claude code for good

if you like claude cowork the open alternative is

GitHub

GitHub - different-ai/openwork: The open-source alternative to Claude Cowork (powered by opencode)

The open-source alternative to Claude Cowork (powered by opencode) - different-ai/openwork

if pi seems overwhelming then opencode is the user friendly alternative

GitHub

pi-vs-claude-code/PI_VS_OPEN_CODE.md at main · disler/pi-vs-claude-code

Comparison between open source PI agent and closed source Claude Code agent - disler/pi-vs-claude-code

cats out. let the llm export wars start

404 | ZeroHedge

ZeroHedge - On a long enough timeline, the survival rate for everyone drops to zero

ai is code, code is speech, ai is speech and in the united states there is freedom of speech. the supreme court will strike down any ban on open source llms

spend time learning to set up PI to ween yourself off of claude code

i'm not sure if LLMs will use bitcoin directly, but I'm almost 100% certain that inference providers hosting unapproved open source models will only accept bitcoin . Just like Silk Road was a boom for bitcoin, black market inference will be a boom for bitcoin .

The total addressable market of OpenAI and Anthropic was just reduced to government partners. adjust your valuations accordingly