x.com/i/lists/19786535179726… use this list and amplify Catholic tweets to make the timeline better.

This requires a pretty low level of humility, honestly. Anyone can get there.

Mathematicians think in proofs, lawyers in constructs, logicians in operators, dancers in movement, artists in impressions, and idiots in labels.

"Idiots think in labels" - Taleb wisdom:

The ancient heresy of Soul Sleep, or “Christian Mortalism.” Mary is indeed fully alive, even more alive than we are, and divinized, being like Christ, as are all Christians in heaven, can easily hear prayers. The prayer of the righteous avails much. She bore God Himself.

We have been thinking about this a lot at Metorial. First of all, MCP is a horribly designed protocol and it took us months to write a platform that fixed all its faults. Our implementation is MCP compatible but not even real MCP. Code mode as a next step makes no sense and it shows that these people have not talked to a single enterprise about their actual needs -- you need permission control, observability, OAuth configs, scopes for particular IP addresses, etc. that can be shared across an organization.

we keep building abstraction layers on top of abstraction layers when LLMs just wanna write code function calling was literally "hey can you return JSON pls?" and now we're writing schemas for the schemas sometimes the simpler solution is just letting it do what it was trained to do

TBH i'm surprised by how slowly/quickly they all seemed to realize MCP was a mistake. There was so much critique of MCP a year ago, but everyone raced to announce support so it effectively became a standard. Unfortunately, none of them were particularly useful so it became kinda easy to dump the standard too.

This is an awesome approach to using MCP servers. We need more people to pay attention to this. Basically, convert a MCP spec into code functions, and have the model write code to call the functions, instead of typical tool-caling. I’ve been doing this with Python for a long time now... you always want to lean into what the model is good at, and obviously, models have seen FAR more code than they have tool calls. Plus, code can orchestrate calls in more complex ways than pure tool-calling can... for example, instead of doing 50 calls to the same tool over and over with different arguments, a model can just write a for loop and do it with way fewer tokens (cheaper, faster, less context bloat!).

imo, there is a major difference between MCP/tool calls and code-mode with code mode control logic can be written once and then eval'd on the client. for multi-step processes it can significantly speed & and simplify. .... BUT that doesn't take away from your main point: "The central thesis of my rant is that OpenAI and Anthropic are platforms and they run "app stores" but they don't take this responsibility and opportunity seriously. And it's been this way for years. The quality bar is so much lower than the rest of the stuff they ship." 💯 My assumptions is that it is a mix of needing to iteratively figure out what it means to have an app store (features, business, integrations, ...) - and a software dev process that doesn't fit When OpenAI launched codex (web) bragging about how it was dramatically changing every day for the last couple weeks (by using the tool) - it showed. The web ui was (and still is) a mess. Many features didn't work if you deviated from the golden path. (a canonical example of vibe coding)

fcking agree 100% ive been thinking of this too yes i can build my own clone, but whats the point? just utilize the reach of chatgpt or claude and let me ship inside as an app been exploring lots of chat apps sdk in chatgpt usecases, so freaking fun

NEW - Israel dumps hundreds of millions into U.S. influence efforts, targeting evangelicals in churches with the "largest U.S. geofencing operation" in history and influencing Google search and AI responses from ChatGPT and Claude. disclose.tv/id/wrbhq1fa5c/

To be honest, I gave it a try, and it's so bad for real projects, I don't get all the hype around it. It simply cannot handle large codebases. It's good for prototyping and doing catchy tiktok stuff like a one-prompt "build me a website", but the more you go, the worse it gets.

I’m using it because it’s free and usually good enough Not because it’s best. Stop with the fake news, your chart is on tokens volume not on it being best

@theo explained this phenomenon clearly about a month ago Grok simply burns more tokens than other models to accomplish the same task, thus any "per token" metric will skew secondary layer to this is it's simply an extremely cheap okay-ish model that is cost effective

I just did a comparison between Grok4, GPT-5 Pro, and Gemini 2.5 Pro Research on a hard coding problem. (How to get mxfp4 working on H100 using only Modular Max and Mojo) Grok 4 was horrible, Grok 3 was better. It hallucinated a bunch of stuff even though it was using web search. I would be embarrassed if I was touting these stats. Half way through the response it started going into commentary about Modular’s mission. Lolol. Gemini 2.5 Pro gave a decent solid answer GPT-5 Pro gave the best answer by a fair amount. I didn’t try Claude Sonnet 4.5 but I expect it would be at least as good as Gemini. But I totally agree on definitely no one should be using Grok 4 for code. I was wondering if xai was buying their own tokens to pad the numbers. Lolol.

Introducing the File Search Tool in the Gemini API, our hosted RAG solution with free storage and free query time embeddings 💾 We are super excited about this new approach and think it will dramatically simplify the path to context aware AI systems, more details in 🧵

Google literally just killed 100s of startups Their new “File Search Tool” (incredibly dumb and misleading name btw) is a hosted RAG solution that allows you to upload files like DOCX and PDF, and chat with them This could be used for things like customer chat bots, where you just put in your company data (FAQs etc) Wild times

This is classic Modernist disingenuity. Vatican II was not a "response" to the cultural revolution, but a deliberate capitulation to it. In the 1960s, when society was in upheaval, it was most necessary for the Church to batten down the hatches (i.e., to double down on sound doctrine, Thomism, the decrees of Trent and Vatican I). Instead, Angelo Roncalli decided to "throw open the windows of the Church to let in some fresh air." Except, unsurprisingly, he didn't let in fresh air - he let in a pestilent cloud of error, sin, and confusion. This is the equivalent of a father seeing an F5 tornado approaching his home, and instead of leading his wife and children to the safety of the basement to weather the storm, he takes them out into the backyard to play on the swing set. Do not be deceived. Responsibility for much of the misery, confusion, and sin present in the world today may be laid directly at the feet of the Vatican II robber council and the false popes who carried out - and continue to carry out - it's impious and heretical decrees.

Comparing autocompletes of 1. Zed 2. Cursor 3. Windsurf 4. Github copilot 5. Augment Code Cursor ~=~ Windsurf > Augment > Zed >>>> Copilot @karpathy which one do you use for yourself?

Fr. Bartholomew Man Jung-hyun was knocked to the ground while distributing Holy Communion during a 2012 Mass in Jeju, South Korea. Despite being beaten by police who broke up the liturgy, he used his remaining strength to gather the fallen Hosts with reverence. The image of this wounded priest—thinking only of the Eucharist—remains a powerful testimony to the sacred dignity of the priesthood.