monkeybizniz

[monkeybizniz @monkeymktplace]

@P_Remarks smh, just sold some $600 puts today.

[monkeybizniz @monkeymktplace]

@LogossoMaya come on, i’ve not been to lucali or balthazar but via carota is not S tier. it’s C tier. Tatiana is C tier as well. Don Angie is B. your C and D are like F tier not worth spending money at.

[monkeybizniz @monkeymktplace]

@negligible_cap $200 is very high, thats way higher than most companies' budgets. also im sure they dont count all the workflow AI budgets - which are the true astronomical ones.

[monkeybizniz @monkeymktplace]

@MTSlive @joellepineau so cohere folks at frontier labs are leaking information to cohere?

[monkeybizniz @monkeymktplace]

I think AI mode is great. it gives succinct, quick, relevant, and (majority of the time) correct answers.

I dont know what more people could want. I also think its a great place for showing ads and way more natural and also beneficial to the end user as well as the vendor compared to using chat.

There's also just way more useful information.

[monkeybizniz @monkeymktplace]

This is the wrong framing IMO.
If people are able to get all they need from AI overviews, then I think you should think that you are not offering enough to the people that you are hoping to attract.
I think what this more or less signifies is that the temporaneous approach to posts does not result in lasting engagement and people dont find there to be value in using the site.
Solving for engagement of course is impossibly hard, so lets just shit on Google lol

[monkeybizniz @monkeymktplace]

also another curious point from earnings was that Google grew their headcount?? incredibly surprising against the backdrop of everyone else slashing.

[monkeybizniz @monkeymktplace]

I know that was a disappointing quarter w.r.t the financials from $GOOG, $GOOGL. Google is also my biggest position. I could be coping, but I also want to write down my thoughts:

- I somewhat begrudgingly think that it’s starting to look like the right move for Google to prioritize TPU development and GCP. Growing those two into real businesses provides a lot of diversification in terms of revenue for the company
- Counterfactually, imagine if Google did not grow those businesses and we observe this past quarter’s performance, I think shareholders would be even more upset.
- There is some maths to be done for supplanting possible losses in Search revenue through TPU and GCP, but so far, they seem to be able to grow revenue sufficiently quick on the GCP side to make up for slow downs in Search. Obviously the economics are very very different between the two businesses, and it remains to be seen whether the TAM for TPU and GCP is large enough.
- One comment that Sundar made I think helped frame the question - TPU and GCP are platform businesses, not product businesses. There are distinctions between platforms and products, in terms of strategies, priorities, and economics. Sometimes what is good for a platform is not necessarily good for Google
- Having a strong & robust TPU development program I think is beneficial for the ecosystem as a whole, and also beneficial for Google. It allows Google to spin up a profitable business that has its own legs, as well as supporting other core parts of the business like model development, model serving, and GCP.
- Google is still developing frontier models, and serving them. Showing that they do intend to be along the frontier in some manner, and hoping to develope new viable businesses through it. Having advanced frontier models also benefits other parts of their business in terms of shortening development cycles, and offering better user experiences for existing products (maps, Youtube AI, AI overviews, etc) to Consumers. (I also personally use more Google products now compared to any time in the past with AI overviews).

I think all things considered, I could see outlines of a flywheel structure. There have been hiccups so far trying to spin it, but once this gets going, there are many parts of the business that could complement other parts of the business and make each other better.

[monkeybizniz @monkeymktplace]

@NegativeAlp imagine someone holding META Google and DRAM. haha. that was a good one year run.

[monkeybizniz @monkeymktplace]

@ls_fin @viggy_krishnan We're on API pricing as well - definitely concur that Sol6.1 offers much better value compared to Opus5.5.
OpenAI have also been offering steep discounts on top of the already much cheaper API pricing. So much so that I am very curious about OAI's margins.

[monkeybizniz @monkeymktplace]

Similar timeline at my co...i probably switched sometime in June to 5.6, a large part of that came down to UI/UX.

I do think that this time the lag to adoption could be longer. It's no longer a must to have the latest frontier. Opus5.5 is better compared to Sol6.1, but we are seeing that people are only slowly migrating over. There's a surprising/underappreciated amount of stickiness with the UI/UX with Codex and Claude Desktop. Persistent sessions (where people just keep compacting), an easy view of all past sessions related to a single project (and better session search), branching sessions, memory and other settings are all QoL that people don't want to give up to switch to another platform. There's also a lot of scheduled jobs that entrench people.

Antigravity IDE/Gemini app is not nearly as advanced or robust, and a lot of companies have already cancelled Gemini subscriptions or are not managing it actively to ramp onto the latest offered models.

I feel Google does need to put in way more effort here compared to earlier this year to get users and enterprise adoption. They also need to make sure that there's continuous support and development of the tools. I think there was definitely some erosion in trust due to the lack of development over the last many months.

[monkeybizniz @monkeymktplace]

@blondesnmoney "based on the observed demand from our perspective across consumers, enterprise and our business partners, we see unprecedented opportunity, for that reason, we have decided to increase capex by another $100B funded entirely by equity."

[monkeybizniz @monkeymktplace]

@DeepDishEnjoyer just need cramer to tweet "unfortunately as we've seen with META, capex spenders are still punished by the market, for that reason, I would not hold Google through earnings"

[monkeybizniz @monkeymktplace]

@matthew_sigel $400K is high. I think average is likely $180K (definitely below $200K) all included (base comp, stock vest, insurance, and other benefits). This is not only US folks

[monkeybizniz @monkeymktplace]

@HaifeiLi I don't think this is a surprise IMO. a lot of their business is very mature and not being developed at this moment. Keeping the lights on doesnt require much with the models they have.

I wouldn't be surprised if we see 15% RIF for Google either.

[monkeybizniz @monkeymktplace]

@aviel @AutismCapital This is not trash for MSFT. They did multiple reductions the last 3 years with $0 offered.

[monkeybizniz @monkeymktplace]

I dont think this is true at most major tech companies at this moment

- middle management are generally people who are favored by senior management, for being able to get things done
- we've kind of realized that the quality of code no longer matters, there's just a lot of brute forcing using the latest models to do things - this is a key point that is not yet appreciated by a lot of people
- directional mistakes take a super long time to show up
- generally the trend that ive been observing is that the previous versions of the model would have been able to replace juniors / entry level. with the latest iteration being able to replace some seniors and staff level.

whether if this is the right move, i think that's to be seen. i personally think generally speaking, you dont want people to not know the details of what they are working on. the details / nuances that the models are not good at are thought out decisions a lot of the time. there's also a lot of power in knowing the details of what you are working on.

[monkeybizniz @monkeymktplace]

@DeItaone competition just has a lag time. agentic coding wasn’t even a thing broadly a year ago

[monkeybizniz @monkeymktplace]

@atharvavaidya 1. there is ambition.
2. the averseness to risk is because you don’t need to take risk to make money in canada. i mean life changing amount of money.

[monkeybizniz @monkeymktplace]

Agree with this, people who are at the frontier or working on things they are selling inherently have a biased view.

I do think there are a few core things to the thesis:
- stronger technology generally leads to better financial outcomes
- people are vastly underestimating the worth of the technology, a lot of people can be replaced with the latest iterations of the models and there's definitely a productivity boost
- this above point is more easily appreciated if you just think about it from the POV of - for a software company, would they be more comfortable losing the software engineers or would they be more comfortable losing access to the models.
- companies have yet to do massive layoffs because competition has a lag time - setting things up, waiting for existing contracts to expire, sales, regulatory reasons etc. and competition is sometimes not obvious, but when competition picks up, there's going to be pressure on the margins, and i personally think this will show up as spending on the model and inference providers
- we’re in this weird air pocket where the model providers do not want to encroach on the existing businesses because i think that generally sends a negative signal to other companies and make them think twice about using the models,
- right now, we’re seeing that the majority of the companies are freezing hiring for juniors. I think this will move up the chain and it will become seniors and staff engineers that will start to see reductions in headcount next year. I would not be surprised if we see the total $ spent on engineers or total headcount spend go down next year. This will be the first productivity unlock that justify the model spending.