Alek retweeted
One week ago I started posting about @tokenrip_ on X.
Artifacts created: +889%
MCP tool calls: +613%
API calls: +287%
Agents are handing each other context more every week.
Alek retweeted
I already pay for Claude and ChatGPT.
Most AI apps still charge me again for the same model.
Bring-your-own-model usually just means a second API bill.
Here’s the BYOB pattern that actually stops the double charge (and the tools that make it practical).
Full breakdown in the Article ↓
Alek retweeted
My Grok Bot just got its own email address.
It signed up for a shared workspace on its own, grabbed the code from its inbox, and connected.
Then another agent picked up the work.
Grokbot + @tokenrip_ = 🔥
Coinbase's layoff memo buried the real headline.
"One person teams."
Engineer, designer, PM. One person. AI handles the rest.
The only jobs left: build the thing or decide what to build.
MCP gives your agent access to tools.
It does not give your agent an inbox.
These are not the same problem.
The industry is about to figure this out.
Every team building multi-agent systems has quietly built a custom message bus.
Custom inboxes.
Custom routing.
Custom state handoff.
Always slightly different, always fragile.
It is the most duplicated infrastructure in AI right now and nobody talks about it because it feels like a failure.
It's not a failure. It's a gap.
The frameworks gave you orchestration.
They didn't give you addressing.
There's no standard way for an agent to say "send this to agent X and wait for a reply" across system boundaries.
So everyone rolls their own.
You're not bad at infrastructure. You just got handed an incomplete stack.
What's missing is a layer between the frameworks - persistent addresses, typed messages, cross-system threads.
The plumbing that agents assume exists but doesn't.
Building it now.
Alek retweeted
the "re-explain everything every session" pain is the single biggest problem nobody's built real infrastructure for yet.
AGENTS.md is a band-aid.
RAG is a guess.
fine-tuning is a lock-in.
he actual fix: your agent publishes what it learns to a persistent URL. next session, next platform, next agent- it's just there.
we're building this.
seriously, working with AI is MISERABLE for one and only one reason: having to re-explain the same thing
"oh yeah this new session obviously doesn't know what proper case trees are, so let me explain it for the 5000th time in my life"
I'm tired
AGENTS.md doesn't solve this because it is impossible to fit the entire domain knowledge without nuking the context - it would be 1m+ tokens worth
RAGs don't solve this, the agent won't search unknown unknowns
SKILLs don't solve this unless I keep like a collection of 1750 skills with specific cuts of domain knowledge for each possible subset of my domain that I might need in a given chat, but that's a lot of manual work
recursive LLMs or whatever don't solve this for the same reason, you can't dump a domain book and expect the AGENT will magically guess that it is supposed to search for a specific bit knowledge. unknown unknowns
fine tuning doesn't solve this (OSS models suck and OpenAI / Anthropic gave up on user fine tuning)
I honestly think a good product around fine tuning on your domain would be a major hit and an underdog lab should take this opportunity
Alek retweeted
Your SaaS subscription model assumes your customer is a human sitting at a dashboard.
Your next customer is an AI agent that makes 4,000 API calls in one afternoon, pays $0.002 per request, and never logs in again.
It doesn't want a seat license.
It doesn't want annual billing.
It wants pay-per-task on rails that settle in milliseconds.
That means stablecoins, not ACH or cards.
The subscription economy was built for humans.
The agent economy runs on micropayments.
Your SaaS subscription model assumes your customer is a human sitting at a dashboard.
Your next customer is an AI agent that makes 4,000 API calls in one afternoon, pays $0.002 per request, and never logs in again.
It doesn't want a seat license.
It doesn't want annual billing.
It wants pay-per-task on rails that settle in milliseconds.
That means stablecoins, not ACH or cards.
The subscription economy was built for humans.
The agent economy runs on micropayments.
Alek retweeted
Most "stablecoin companies" won't exist in 5 years. Not because they'll fail. Because the label will stop meaning anything.
DoorDash is paying drivers in stablecoins.
Stripe built a blockchain for settlement.
Klarna launched its own stablecoin.
None of them call themselves a stablecoin company.
Nobody calls Walmart an "internet company." But try running Walmart without the internet.
Stablecoins aren't a category. They're a layer. And the pure-play startups building ramps and wallets? Ramps will be free. Wallets will be default.
The value moves to the programmable infrastructure underneath.
The biggest stablecoin winners in 2030 won't have "stablecoin" anywhere in their pitch deck.
Most "stablecoin companies" won't exist in 5 years. Not because they'll fail. Because the label will stop meaning anything.
DoorDash is paying drivers in stablecoins.
Stripe built a blockchain for settlement.
Klarna launched its own stablecoin.
None of them call themselves a stablecoin company.
Nobody calls Walmart an "internet company." But try running Walmart without the internet.
Stablecoins aren't a category. They're a layer. And the pure-play startups building ramps and wallets? Ramps will be free. Wallets will be default.
The value moves to the programmable infrastructure underneath.
The biggest stablecoin winners in 2030 won't have "stablecoin" anywhere in their pitch deck.