deepfates

Based on 0 decided calls, 0% coverage, data through 2026-07-23, there is simply no verified track record to weigh. The ANTE engine files deepfates in its 'Building' band with insufficient corroboration — zero tracked positions, no wins, no losses, no profit-and-loss field, no instrument mix. That is not a knock; it reflects what the account is. Nothing in the 200-tweet corpus reads as a datable market call, so any record-based verdict would be pure invention. Treat the score here as unwritten, not as a trader who failed.

What a follower actually receives is a running feed on how frontier models behave — agent harnesses, self-modifying memory, multi-agent loops, alignment epistemics — not entries, targets, or stops. The register is narrative and conversational, threaded through replies rather than posted as standalone theses. When a claim carries a 'why', it points to papers or lived experiment ('LLM internal states are being studied'), which is more rigor than most timelines offer. But there is no plan format to trade here, and no market horizon that could decay.

This card usually measures how a creator treats a losing trade — stop discipline, loss magnitude, whether losers get posted or quietly vanish. None of it applies: the pack shows no positions, no target-and-stop rate, no average win or loss, and no scoring anomalies. There is nothing here that can preserve or vaporize capital, because nothing here is a capital call. The intellectual risk is a different animal — strong, sometimes combative priors on machine consciousness — but that is a debate to join, not a drawdown to survive.

This is where deepfates lives. The voice is high-volume shitpost-philosophy: memes about ascending to 'the Beyond', deadpan jabs ('skill issue probably'), naturalist asides about ravens versus crows, and real curiosity about what agents 'want'. He owns his own biases openly — 'I'm also resistant to other people's narratives' — and needles interlocutors without sneering, though the sarcasm runs hot mid-argument. No Discord, newsletter, or masterclass funnel surfaces anywhere in the pack; the honest monetisation read here is simply none evident.

There is real teaching in the mix: modelling theory-of-mind for a language model, distinguishing low- versus high-thinking behavior, pushing others toward 'controlled and reproducible settings' and actual research over anecdote. He credits collaborators and points followers to papers and producers he rates. The community posture is generous to peers but sharp with sloppy arguments — 'you have to try harder' — so a thin-skinned beginner may bruise. For the AI-curious willing to keep pace with the jargon, the thread reads like a seminar.

You'll like this if you want frontier commentary on how today's models actually behave, enjoy dense reply-thread discourse, and don't mind philosophy delivered as shitpost. You'll like it if 'the intelligence is stored in the society' reads as a feature, not a riddle. Skip it if you came to ANTE for datable market calls, entries and stops, or a track record you can grade — this account offers none of that, and never claims to. It is a mind to follow, not a signal to trade.