Alvin

my mind feels clearer and more at peace with 3 mythos-tier models at the frontier. multipolarity is all but guaranteed. now we just need k3 base so we can make a unique post-train without any distilled tokens. free range kimi

you awake from a deep trance, step away from the phone to see your friends & family

@brivael The survival probability of firms who maintain a significant short position in SpaceX over time is very low

Bad, bad, bad. Opposite of the reaction needed from the USG in light of the Chinese gov't deciding to go the other direction. Not exactly the type of stuff you expect from an admin that preaches about "accelerating innovation" and "free markets"

@bubbleboi It’ll be fucking amazing, cheap tokens for everyone

My guess is the cost is because Kimi doesn’t have enough infra to serve it plus they are just charging the market price + some discount for a model that rivals Opus. I strongly suspect once the weights are actually open the cost per million output tokens will reach $5.

Markets price the future. Charts of the Week: a16z.news/p/charts-of-th…

The result of the Kimi K3 release will be: less regulation, more capital expenditure, faster releases. In a word: acceleration.

The biggest AI adopters are increasing entry-level hiring. Charts of the Week: a16z.news/p/charts-of-th…

Let me show you how to run Kimi-K3 locally, right from your own house

@bubbleboi Your hedging must be insane, or maybe just diversified portfolio You should go all in on $INTC

I’m down only 1% today

@synthwavedd Just gotta wait until they open source it, maybe even API price competition then

lmao moonshot do NOT have enough capacity to serve K3 man. gonna come back tomorrow and pray they've recovered a bit but throughput even when requests succeed is pretty bad

insane. they distilled american frontier models that haven't even been invented yet. damn chinese

@jimcramer So bullish?

tough call remembering that Western Dig fiasco

Straight facts

TSMC Q2: FY26 growth and capex move higher; $TSM -4% pre-market on the 3Q margin bar $TSM $402, -4% pre-market. TSMC printed Q2 revenue of $40.2B, at the top of its $39.0B-$40.2B guide, and $4.31 ADR EPS versus $3.81 Street. Revenue grew 36% y/y, while net income grew 77% y/y to NT$706.6B. The company raised FY26 USD revenue growth to slightly above 40%, from above 30% previously, and raised FY26 capex to $60B-$64B from $52B-$56B. The $62B midpoint is $8B, or 15%, above the prior midpoint. Q3 revenue guidance is $44.6B-$45.8B, or $45.2B at the midpoint. That is another $5.0B, or 12%, q/q step-up from Q2 and above the $43.7B Street view. The stock reaction is centered on margins. Q2 gross margin reached 67.7% and operating margin 60.3%, both above guidance. Q3 guidance moves to 65%-67% gross margin and 56%-58% operating margin; the 57% operating-margin midpoint is below Street. After a 77% y/y gain in net income, the market is asking how much overseas-fab cost and N2 ramp dilution offset the higher revenue base. Leading-edge mix remains the supporting data point. 2nm represented 3% of Q2 wafer revenue, 3nm 30%, 5nm 33%, and 7nm 11%; 77% of wafer revenue came from 7nm and below. Management expects a steep 2nm ramp in Q3, which supports the next leg of wafer revenue but also keeps mix, yield and depreciation central to the margin discussion. Takeaways 1/ The $8B capex increase supports a larger FY26 leading-edge and advanced-packaging build. $NVDA, $AMD and $AVGO demand remains visible at the foundry level; $ASML and the semi-cap group get a clearer equipment-spend backdrop. 2/ The Q3 revenue midpoint puts TSMC on a $180B annualized revenue run-rate before any further Q4 step-up. FY26 growth above 40% now requires H2 revenue to remain materially above H1's $76.1B. 3/ Margin is the next estimate variable. A 100bp move in gross margin on a $45.2B quarter is roughly $452M of quarterly gross profit, before tax. That is why the Q3 margin midpoint matters more to the tape than the Q2 beat. TSMC also committed an additional $100B to U.S. manufacturing, including four more Arizona fabs, bringing its stated U.S. commitment to about $265B. That is a multi-year capacity and supply-security program; it should not be read as FY26 capex. Source: TSMC 2Q26 earnings release, presentation and conference call.

what an absolute huge win for open source AI in america. 1st ever model from OpenAI's ex CTO and its a 1T open model that reasons across text, code, images, video - but the best part isn't even how good it is... its how you can fine-tune it to create something way better: > you can adjust the thinking effort > you can influence the way it approaches a coding problem > you can perfect token spending so your task costs pennies this solves a big problem that companies are facing now: they're spending $100Ks - $10Ms when they could just be fine-tuning a "less intelligent" model to do the same task at a fraction of the cost (and often better) thats why bridgewater's using Thinking Machine's tinker product i expect to see a lot of folks playing around with this model, finally an alternative to chinese open source!

Then sun is relentless this summer, and so is Sol in Codex

I know many of you think it's cringe when I bullpost $NBIS on red days. I politely have to say I think the opposite. When the stock rallies like crazy and everyone who was saying "this shit is going to $100" starts acting bullish, that's cringe to me. It's dishonest, weak and shows herd mentality. I bullpost on days where the business makes fundamentally great progress and the stock goes down, because those are the days where the risk-reward gets so much better. When something good happens and the stock goes up, it's pricing it in. Stock down + great news is the actual asymmetrical opportunity everyone on X always talks about, but when you get these opportunities half the people here start shaking and sell close to the bottom where you're supposed to buy. Not giving advice or anything, but this is the difference between the people with good performance and bad ones. The buy high sell low meme is so real and human history remains undefeated. We see this shit every few months over and over and over again. If you want to draw a lesson from this it's calculate and act based on your estimates, not emotions.

Good to see downside being priced-in into memory stocks, and is needed/healthy from a technical perspective. The next leg higher (we think $MU to $2k) will be the de-risk of what a bottom could be, ie higher than expected gross margins as Y/Y growth decelerates modestly.