Marky Thompson

Stocks

[Rittenhouse Research @RHouseResearch]

Burry’s bear thesis on NVIDIA $NVDA went from:

“The accounting for GPUs overstates their useful lives!”

to:

“I have assessed this pre-revenue chip startup to have exceptional technical and engineering capabilities, and believe they represent a significant threat to NVIDIA”

real quick.. Lol.

[Marky Thompson @twotter3141]

@brentread_7 @codesportsau Over the last year, @brentread_7 has been head and shoulders above the rest of the press pack with his takes.

[Marky Thompson @twotter3141]

@Ojeda4America They are the dumbasses that voted for this

[Marky Thompson @twotter3141]

@fammetaX Retard

[Marky Thompson @twotter3141]

@BrianLaManna_ How are alll these fintwit dumbasses somehow discountbging the fact that it’s part of cleaning house and removing the shit performers, not people happily leaving of their own volition?

[Marky Thompson @twotter3141]

@puckrin The fact you compared gdp with market cap means you a) have a net worth of $10 and b) have 0 understanding of finance.

[Marky Thompson @twotter3141]

@michaeljburry @amazon That’s a bit silly mate, huge false equivalency.

[Jeremy Raper @puppyeh1]

Live look at Quadrant selling a Kiwi media shitco that’s been unsellable for 2yrs to $SEG.AX at 5x EV/ebitda when every other listed Aussie media radio shitcos trade at 2.5x 🤡

[Marky Thompson @twotter3141]

@TMFScottP From the PBO’s report: “Across the middle 50 per cent of the distribution, the average
tax rate increased from 6.6 per cent at the low end to approximately 13.0 per cent at the high end.” Literally a doubling of effective taxation for middle band income earners. “Small impact”…

[Marky Thompson @twotter3141]

@TMFScottP “Would not be a crisis for the vast bulk of people” is a specific scope identification phrase that you’ve used to inherently exclude/ignore bracket creep and its affect on the “vast bulk of people”.

[auonsson @auonsson]

The yacht of Mark Zuckerberg allegedly refused repeated calls by US Coast Guard looking for ships available to rescue a small boat with 2 women, a kid and a dog.

According to a user on BSKY who was on a cruise ship that performed the rescue eventually. AIS-track fits claim.

[Marky Thompson @twotter3141]

@HedgeyeComm 0 attractiveness to enterprises as there’s no training opt-out. It will remain a joke until they change that.

[Trung Phan @TrungTPhan]

Charlie Munger’s three Ls is undefeated

[CAMERONWILSON @cameronwilson]

"we do not want to highlight this yet"

Hyped data centre company Firmus seemingly accidentally reveals plan to source water from a publicly owned provider that almost exclusively serves farmers, thanks to an inadvertent note left in its website FAQs lol

[Jukan @jukan05]

BofA’s Vivek Arya made an excellent point, which I’d like to share.

Is open-source AI bearish for memory?

Closed models amortize global demand across shared HBM pools concentrated in a handful of data centers, whereas open models create a new memory footprint with every deployment. If 10,000 companies self-host the same open model, the model weights must be replicated across 10,000 separate HBM pools, with each deployment also requiring its own KV cache. As 128K–1M token contexts become commonplace in 2026, the KV cache alone can exceed 40GB per active session.

Low-cost Chinese APIs drive greater inference demand, broader enterprise self-hosting, and more memory sockets worldwide. In short, closed models concentrate memory demand, while open models multiply it.

[Trade Whisperer @TradexWhisperer]

Kimi K3: same intelligence tier, same cost tier as GPT 5.6 Sol Max.

No efficiency breakthrough. No cost collapse. Just another frontier model burning the same compute and the same memory as everyone else.

$MU $SKHY $DRAM

[Zephyr @zephyr_z9]

K3-type models are bad for frontier AI model companies and margins (K3 won't threaten OAI, cuz Sol is still very price competitive, but Opus will be evaporated) but they are great for cloud/TaaS (token-as-a-service) providers and infra
Moonshot cannot keep up with the demand, and they will invest a lot on acquiring compute (good for A shares AI infra plays)

[Marky Thompson @twotter3141]

@kimmonismus Lmao like the Chinese can be trusted without verification.

[Marky Thompson @twotter3141]

@alexandr_wang Can you give an options to not train the model off inputs?