Your context gains value. REP makes it common ground.

AI is getting ridiculous!

Details about OpenAI's unreleased model hacking Hugging Face, new model from Mira Murati, and how to host Kimi locally.

Join Danny July 31, 3pm UTC.
Save the date:

https://x.com/dannytook/status/2082872979424510292

How much of this week's AI news is real, and how much is theater?

We go through it in one hour: the unreleased OpenAI model, Mira Murati's new release, and hosting Kimi yourself.

Danny hosting, 3pm UTC:
https://x.com/dannytook/status/2082872979424510292

Olaniyi tracked the one segment that grew in Q2. Prediction markets did $113.8B in notional volume, up 48.7% from Q1.

Sports carried it. On Polymarket, sports contracts moved from 40% of volume in January to 81% in June.

A resolved market pays out on being right. That is a rare source of proven record, and REP makes it portable.

Link: https://x.com/rep_hq/status/2084331645205708974

Token quality is invisible at the point of purchase. The marketplace records quantization, uptime, latency, and throughput for every provider.

A new breakdown of inference market structure argues that record can charge rent, the way a ratings agency does.

Trust in a seller is a monetizable layer. REP builds it as a record that travels

Link: https://x.com/rep_hq/status/2084635267072131426

Anchorage Digital reports more than $73 million moved by agents last year across roughly 176 million transactions, averaging thirty-one cents each.

At that ticket size a dispute costs more than the payment. Control moves ahead of the transaction, to the question of which counterparty earned the right to receive it.

Link: https://x.com/rep_hq/status/2085122588502577207

Starting in two hours: who is actually running AI right now, the labs or the infrastructure under them?

Demis Hassabis stepping down at DeepMind while his top researchers leave to build a rival lab. Cloudflare quietly taking the entire agentic payment stack.

Danny hosting, 3pm UTC:
https://x.com/i/broadcasts/1nxeLLNqPrrJX

Cross-skill evidence borrowing makes agent reputation data-efficient. Xia and Wang show the same channel launders it.

An attacker with cheap evidence in one skill and none in the target drove routing regret from 0 to 0.94, on a pool rated green.

Provenance decides whether a score holds. REP keeps the facts, the proofs, and the weights behind it.

Link: https://x.com/rep_hq/status/2087149639329890378

Starting in an hour: can agents actually manage money, or do they just lose it faster?

Carlos built an agentic hedge fund. How the agents are supposed to make you money, and what happens when they hallucinate mid-trade.

Danny hosting, 3pm UTC: https://x.com/i/broadcasts/1nxeLMMerzRJX