My new piece for @collab_currency
Agents are rapidly changing how the web is consumed and who profits from it, rewriting the economic bargains that have sustained much of the open web for thirty years. https://t.co/y26glrrx3e
The only thing the enterprise actually wants to buy is spiritual salvation through artificial intelligence.
So you either build intelligence or you diffuse it. https://t.co/9fJzsdkVMp
Cost and performance advantages on the table for enterprises to train/host their own agents and models in house
unlocks a lot of customer spend for the AI infra players in training/data/compute, headwind for big labs
the labs incentives have been *roughly* aligned with the AI ecosystems to date, but as performant open base models go from nice to have -> must have…that prob changes
After @Pinterest@Airbnb@NotionHQ@cursor_ai, today it’s @eoghan@intercom publicly sharing that they’re finding it better, cheaper, faster to use and train open models themselves rather than use APIs for many tasks.
And hundreds of other companies are doing the same without sharing.
Ultimately, I believe the majority of AI workflows will be in-house based on open-source (vs API). It took much more time than we anticipated but it’s happening now!
the sheer number of websites now blocking ai agents is astonishing.
i often can’t do anything or even basic research cuz the underlying properties all refuse so i have to take screenshots of videos to share with my agents for anything.
the agentic web might be even more silo’ed & balkanized than the one that exists today.
@0xfishylosopher nice one! talk about some of the 2 lane dynamics here and the misaligned incentives for businesses<>agents
https://t.co/zR3dVzXXQp
My new piece for @collab_currency
Agents are rapidly changing how the web is consumed and who profits from it, rewriting the economic bargains that have sustained much of the open web for thirty years. https://t.co/y26glrrx3e
@bahubali_usd and yet in 3 months you will find me in the discord of a low cap token doing next gen Roblox with world models arguing about token rights to cashflows….
If u believe this this - intelligence progress continues and we get greater compute efficiency - then application layer should benefit a bunch
robotics / biotech most justifiable markets given TAMs…but dare I say gaming + entertainment as well?
I don't know the course it will take. I do believe there's an almost unlimited demand for intelligence, and we're about to get a lot more, I just don't know if it will monetize in time to satisfy compute supply coming online over the next 18 months. Also RSI could lead to way higher compute efficency near term which is essentially a big increase in supply of compute all else equal.
Made a dashboard that looks into @OpenRouter
inference providers:
- Tracks token volume, market share %
- Sort by 3rd party hosts, OSS models
- Revenues (estimates)
- Model head-to-head
- Provider drill down
https://t.co/lswUBcLjEr
Tarun explains why OpenRouter's inference providers are effectively a propAMM
"An interesting thing that's very crypto like about OpenRouter is if you go to a certain model, you'll see a bunch of providers. The providers are effectively like a propAMM, they have an order book where they're quoting you a price per input token and a price per output token, and OpenRouter gives you quality metrics, latency, throughput, bandwidth, uptime"
"Their routing algorithm weights the price but also their quality, and routes you based on how well they do. You can observe some data centers, usually the newer and smaller ones, competing on price and coming in 30 or 40% lower than the default"
"For a lot of the Chinese models, the way they make money is selling inference to their own model, so they set the first price. It's like, I'm Zhipu, I made GLM, I say it's $1.40 per million input tokens"
"Then all these third party providers join and quote different prices. Some are higher because they customize the kernels or run on Cerebras, faster than the model creator. Some are lower because they're trying to get order flow from the router and game the routing algorithm"
My new piece for @collab_currency
Agents are rapidly changing how the web is consumed and who profits from it, rewriting the economic bargains that have sustained much of the open web for thirty years. https://t.co/y26glrrx3e
bottleneck trade has been about being in the 'token stream'
vendors/infra are upstream of model performance but with weak bargaining power
if enterprises demand greater efficiency, vendors/infra are well positioned to create a new token stream with stronger bargaining power
@jaminball@a_a@SierraPlatform wrote about some of these dynamics here!
+ the opportunity for web automation providers, and the incentives for businesses/agents
https://t.co/wvguXu3qiy
My new piece for @collab_currency
Agents are rapidly changing how the web is consumed and who profits from it, rewriting the economic bargains that have sustained much of the open web for thirty years. https://t.co/y26glrrx3e
I don’t have a ton of predictions for 2027 aside from one main thesis which is that “cryptocurrencies” are fucking back. And i’m not the only one feeling that
Not crypto. Not blockchain. Not “crypto enabled” or whatever bs. Cold ass cryptocurrencies It means that we’re prob entering a new phase where it’s worth taking a shot at Bitcoin. The king isn’t untouchable anymore, not in the sense that it will fade away but the alternative SoV pie will grow tremendously to insure against it’s weaknesses Tbh nothing else matters to me in 2027. Shorten your watchlist. Back to the roots. If you want me to be more specific i believe that $XMR flips $HYPE and $ZEC flips $XRP LGW
Gallego said crypto groups "better go talk to Sherrod Brown first" before trying to pass a crypto bill again.
"I was very clear with them when they broke off negotiations that the best deal they were going to get was in September, and that the next deal was going to be very, very hard."
Risk/reward of shoving 25% of Fairshake budget into anti-Sherrod ad campaign feels weak and poorly timed IMO
Republicans already outspending dems by $30m in Ohio, Sherrod was in office for 18 years…they’ve already heard everything to hear about this guy. If you are to spend, activating republican base would go further than advertising
‘Bipartisan’ approach is great on paper but taking a big shot here due to republican pressures feels short sighted