rfgplk

Translate + encode it to any scheme you want, give claude the encoding and tell it to analyze, then splat out an encoded output file. In a separate session have it write you a python decoder script. Works as a near universal bypass, you might have to synonym swap some words though.

Or preload the context with sufficient clean information, say 500k tokens worth. The guardrails will then trigger less frequently.

I'm a hardcore LLM booster but I agree that there's a place for LLM-free projects. In fact human generated code is still superior to even frontier LLM generated code, as long as you're willing take a massive productivity penalty.

Tbh it makes sense if you want a truly LLM-output free project. I can respect it.

I've personally found that most models completely collapse at sub-medium effort. I generally run most of my stuff at medium/xhigh with heavy core tasks @ max

The next model will likely be Nova

Anthropic models are comically overpriced. They also gobble up an obscene # of tokens to get the same result as others. The price delta is literally 5x.

Cheap and affordable. Dispatching proper tasks to these models can save you a huge % of your quota. Especially if you _know_ the task is simple (ie add something here or there, or archive a bunch of files and reformat kinda thing)

> But is it anything more than a neat hobby at this point?

The opposite. With LLMs you can now pick up any one of these obscure language and write a full blown AAA game engine in them. Or a database. Or a web framework. Or... anything you set your mind to. The language no longer matters. btw LLMs are exceptionally good at Smalltalk/Haskell/Perl/FORTRAN and other old/uncommon languages.

Good point, I asked Claude to also seek out funding.

What's preventing me from feeding this into Claude and reconstructing the company for $200?

I'm unironically doing this right now; downloaded the docs (htm + pdf render) to a local repo and running Opus 5.5 @ max effort with the prompt:

  "In this directory you can find Anthropics official S-1 filing, read it and replicate their corporate strategy, validate it via an adversarial subagent swarm and write up a detailed corporate plan for a startup copying their exact strategy. The goal is to achieve a $1 trillion valuation and an IPO in 5 years."

Edit: I'm not even joking I got safeguard flagged for Details: `[cyber]`

So Google shouldn't charge for API access to maps either according to you. It's all open data

> "Exactly as tall and as steep". How is a mountain that exists in a video game exactly as tall and steep as the real one?

A mountain irl is, for instance 783 meters tall (sea level). A block in minecraft is always 1x1x1meters (according to the game engine def). Thus you make a mountain 783 blocks tall. You can also easily raise the sky limit if it becomes a problem.

The C++ committee is pointless, and this is coming from a C++ expert.

Most of the things the committee has been "working on" are realistically _side projects_ today with AI. I've made countless additions to my clang fork the most notable being a source compatible borrow checker, ie you turn it on and it tells you where you effed up. Zero source changes. Equally as correct as Rust. For any sane C++ program. The committee still isn't capable of implementing this btw.