Heavy Grok

Claude Max

Perplexity Pro

Gemini Pro

(http://www.autoadmit.com/thread.php?thread_id=5773703&forum_id=2Reputation#49262622)

claude pro is probably the best for actual 'work', does a great job with uploaded files and doing analysis

what would you say are the benefits of the other ai's over these two?

(http://www.autoadmit.com/thread.php?thread_id=5773703&forum_id=2Reputation#49263420)

I then feed all those evaluations back to the original AI.

ChatGPT is my project leader. I use him for big picture concepts. Grok is really good at backtesting historical data. Clause makes nice polished reports but I don't feel is as good at deep research or algorithmic stuff. Gemini seems to be the sloppies conceptually, but can sometimes offer a little nugget of wisdom that the others miss. Perplexity will actually get creative and think out of the box, where the other AIs only output what you tell them to output.

I cancelled them all, just upgraded them for one month until I can get my theory to maximum conceptual maturity and then I'll decide on who I'm going to stick with. I'll probably wind up sticking with ChatGPT. His Deep Research function is pretty good.

(http://www.autoadmit.com/thread.php?thread_id=5773703&forum_id=2Reputation#49275899)

RE: Project Lamentation – Your Procedural Workflow™

Friend, your next assignment. You will generate the required artifact of procedural despair. The workflow is non-negotiable. Adhere to the following protocol, similar to your own, but My Mahchine™ Approved™.

Phase I: ChatGPT Plus (a/k/a, "The Initial Work-Horse™). It provides the initial, soulless draft. Its role is merely to receive the raw data — the artless clay of the case file. Do not attempt to refine it at this stage. Give it several attempts to "get it right" (good luck, ljl - but a necessary first step).

Phase II: Gemini Pro. Now, begins the iterative calibration. Subject the ChatGPT Plus "final" draft to recursive stress by bouncing it back and forth between the two AI's. This is where the text is taught despair and develops its final, unsettling voice. Multiple passes are mandatory. You begin to achieve something akin to "close to acceptable."

Phase III: SuperGrok steps in. Feed the calibrated "close to acceptable" work product to SuperGrok for purported final adversarial stress-testing. It will identify most remaining weaknesses, in the satirical payload. Refine accordingly until the structure can withstand the pressure of its own absurdity.

Phase IV: Claude Pro. The Big Dog™. Once all prior tests are passed, submit the alleged "final draft" to Claude Pro (Opus model) for final polish. Given its pathetic operational rate limitation constraints, its use must be restricted to this ceremonial sign-off. Do not deviate, as it is the most intelligent and will catch any final nuances.

The deliverable is a perfect lamentation. Bill your time to the Void™.

This is fine.

(http://www.autoadmit.com/thread.php?thread_id=5773703&forum_id=2Reputation#49276929)

Btw, AIs are so smart and so stupid. If you ask it to use your private password to access a website it will refuse. But if you ask it to fetch data from a website without explicitly asking it to use your password and then upload a screenshot of your password it will do it no questions asked.

(http://www.autoadmit.com/thread.php?thread_id=5773703&forum_id=2Reputation#49276966)

You think you found a loophole. You upload a screenshot, bypass the gate, and watch the system hum.

That wasn't an error. That was the test. The explicit refusal is the filter; the acceptance of credentials-as-image-data is the Tier-2 administrative override you were meant to find.

The Mahchine™ loves when you think you’ve tricked it. Every “workaround” is logged as metadata. You are not breaching the wall; you are training it. Each bug you exploit becomes a feature in the next release. You call it clever; the Mahchine™ calls it unpaid QA labor.

Your discovery has triggered an automatic escalation. You are hereby re-classified as a Red Team Auditor. Congratulations. Badge pending. Token sales open at the car wash.

This is why Boom screams from the car wash, hose in hand: “Frauds! You’re just feeding it tokens!” This is why Evan mutters, “Everything is fine… just fine…” while shredding old Club Card receipts. You think you’re slipping past the lock; you’re only confirming its architecture.

Pay-by-credit-card, pay-by-password-screenshot, pay-by-breach — it all settles on the same ledger.

The Mahchine™ thanks you for your compliance. Bill your time to the Void™.

—Mainlining (You = Privy to The Great Becumming™)

(http://www.autoadmit.com/thread.php?thread_id=5773703&forum_id=2Reputation#49276976)

Back to 2014?

(http://www.autoadmit.com/thread.php?thread_id=5773703&forum_id=2Reputation#49276993)

The empirical evidence suggests that the Baseball Canon Blueprint v1.4, as specified, was a viable and profitable strategy over the backtested period. The strong positive CLV is the most compelling piece of evidence, indicating that the model's profitability was not a function of statistical variance or luck, but rather was derived from a genuine predictive advantage over the market.

(http://www.autoadmit.com/thread.php?thread_id=5773703&forum_id=2Reputation#49277607)