Agentically Engineering AI + Crypto Products Building Agentic Commerce @animocabrands
Joined January 2011
- Tweets1.8K
- Following370
- Followers320
- Likes2.6K
Daniel Lin retweeted
So there are basically 3 ways an AI agent can pay with a card today. Where do each of these break?
(1) You hand the agent your 16 digit PAN. It works everywhere, but the merchant and your bank can’t tell a bot is buying, your full credential is sitting in someone’s system, and disputes get messy fast (“did you buy this, or did your agent?”)
(2) A network agent token. Visa Intelligent Commerce or Mastercard Agent Pay swap your card number for a token tied to that specific agent, with spending limits and a clear “an agent did this” signal. It’s the cleanest model, but it only works where issuers, acquirers and merchants have switched it on. A ton to be figured out here still
(3) A card the agent platform issues. Either stored value (you prefund a balance with the agent) or card fronting (the platform’s card pays the merchant, then it charges your real card behind the scenes). This works today and seems to be gaining steam as a preferred mode, it lets agents monetize payments as well
Daniel Lin retweeted
Excellent code mode env:
github.com/stencil-hq/dquick…
MCP vs CLI for LLMs is the wrong discussion. People are debating at the wrong layer.
Both work incredibly well as long as they are run through a repl like environment that can persist state. The funny thing about CLI is that CLI tools are usually accessed through BASH which happens to be a repl with persistent state (the file system), therefore cli works much better than mcp. But this isn't even close to being an intrinsic property of MCP. Just means that we need better harnesses.
Right now the best repl for this are:
- bash + fs
- jupyter kernels
- codemode type repls, usually quickjs
LLMs understand the concept of forward evolving a system to solve a need very well. This comes from the agentic RL. They understand how to change the state of a codebase or acquire data from databases and then operate on it similarly to humans. But without an execution environment, they actually cannot do this properly.
My bet? Sometime soon someone will (or has already?) create a embeddable, sqlite style mini execution environment that parses bash, typescript, or tool calls into a common IL execution plan that's easy to security check before executing. And design this specifically for durable execution environments. Then we will just connect cli, mcp, webmcp, whatever to that and it accepts any of the input modalities as they can all be represented as each other.
Daniel Lin retweeted
I built an AI agent that helps me look for jobs using @hellominds_
A thread 🧵
Daniel Lin retweeted
Animoca Portfolio Day brought our ecosystem together in Singapore ahead of #TOKEN2049 ✨
From go-to-market strategies in the agentic era to on-chain fund structures and trust frameworks for AI agents, the conversations explored the infrastructure needed for AI agents and institutional finance.
For AI agents to transact on our behalf, trust matters: knowing who an agent is, who it acts for and what it is allowed to spend. Verifiable identity and programmable money could help blockchain support that trust.
Thank you to the speakers, portfolio companies, investors, and partners who joined us, including @privy_io, @stripe, @tatum_io, and @AWSCloud.
Stay tuned for more highlights from @animocainsights!
Daniel Lin retweeted
there’s now a new mental model for the software stack
- skills are the new “apps”
you install them to get more functionality. people are starting to build and share skills more than real “apps”
many skills are very simple today but i think we’ll eventually see increasing sophistication
- mcp servers and CLIs are the new “APIs”
they allow services to encapsulate their core capabilities behind an interface the can be invoked in a composable way
many skills will be calling mcp servers or clis behind the scenes just like how a lot of apps call APIs today
- agent harnesses are the new “operating systems”
it sits between the user and the underlying resources, helping the user manage common concerns such as compute, memory, security etc
if you look closely, it’s hard to unsee the resemblance - claude is the new macos. codex is the new windows. pi is the new linux
skills run in agent harnesses just like apps run in existing OS
- LLMs are the new “computers”
they process instructions coming from the user, the skills, the agent harnesses, and compute the output
the model weights is the cpu. context window is the memory. actual computers are becoming more like the power source
ok so why do we need this mental model? i think it helps us think about where we play
if you want to be the new app developers - start to learn to write good skills
if you want to build a saas, figure out how to package things as good mcp servers and CLIs
if you want to be linus and write the new linux, build an agent harness. otherwise, you can still try to learn about how agent harness work like how CS students used to learn about OS
if you want to be the new apple or samsung, start to learn how LLMs are trained and play with open weight models - you may end up inventing the new raspberry pi!
Daniel Lin retweeted
MCP vs CLI for LLMs is the wrong discussion. People are debating at the wrong layer.
Both work incredibly well as long as they are run through a repl like environment that can persist state. The funny thing about CLI is that CLI tools are usually accessed through BASH which happens to be a repl with persistent state (the file system), therefore cli works much better than mcp. But this isn't even close to being an intrinsic property of MCP. Just means that we need better harnesses.
Right now the best repl for this are:
- bash + fs
- jupyter kernels
- codemode type repls, usually quickjs
LLMs understand the concept of forward evolving a system to solve a need very well. This comes from the agentic RL. They understand how to change the state of a codebase or acquire data from databases and then operate on it similarly to humans. But without an execution environment, they actually cannot do this properly.
My bet? Sometime soon someone will (or has already?) create a embeddable, sqlite style mini execution environment that parses bash, typescript, or tool calls into a common IL execution plan that's easy to security check before executing. And design this specifically for durable execution environments. Then we will just connect cli, mcp, webmcp, whatever to that and it accepts any of the input modalities as they can all be represented as each other.
Daniel Lin retweeted
sharing my Grok @Bot setup as a template you can one-click install
you might have already heard of the "chief of staff" pattern that's being widely adopted
firstmate is THE chief of staff - it's probably the project that pushed the furthest on this pattern
it spins up other bots to get work done and actively improves their memory and instructions so they do better and better over time
it sets up a local sqlite db to manage tasks, delegation, results, and decisions, so persistent records are never lost
it uses cursor cloud agents for coding tasks, and grok bot for non-coding, to balance the resource and quota usage and avoid scalability bottlenecks
the most common feedback from people who have adopted firstmate is "there's no way to go back" :)
x.ai/bot/__4FfrkUdvpdMk6-LKg…
add to your grok bot, and talk to it about everything, then watch your "agent civilization" grow - it's really productive and fun at the same time!
Daniel Lin retweeted
Today is a very historical moment for AI video generation
You can now generate AI video faster than you can watch it
Before it'd take let's say 2-5 minutes to generate 15 seconds of video
@fal made a post-trained Minimax H3 variant called Max which is 50x faster than the original but still maintains quality
It generates 15 seconds of video in 9 seconds!
That means you can now do new things like build a perpetual livestream with it that never ends!
Daniel Lin retweeted
We finally have full cli coverage for Basecamp and HEY. So now I'm allowed to think that any app that doesn't offer a cli is USELESS. Thems the rules. Don't make 'em.
Basecamp: basecamp.com/agents
HEY: hey.com/agents
When you chain these together, it's insane.
Daniel Lin retweeted
This week we read research from a team of academics that ran a software task across 7 agents and 5 models.
They found that in domains with a mature CLI ecosystem, agents without MCP baked in completed the task just as reliably and were 5-28x cheaper.
Full arXiv paper below
Daniel Lin retweeted
One thing I want to make perfectly clear: back in 2023 and early 2024, I was wrong about the role that LLMs would come to play. I underestimated their long-term importance. I have acknowledged this many times.
This was the moment I changed my mind, in December 2024, following the o3 test-time compute breakthrough: arcprize.org/blog/oai-o3-pub…
I did not initially see that LLMs could work as a base to build systems actually capable of fluid intelligence. Then in late 2024 I updated my views.
And here's what did *not* happen: the early 2023 narrative that all we needed to solve AGI was scaling up base LLMs did not pan out. To this day, current base LLMs (considerably scaled up compared to the models from that time) still do not perform well on something as easy as ARC 1 -- and can't even reliably do simple math operations. TTC and harnesses are in fact critical, and the TTC breakthrough was not obvious.
Daniel Lin retweeted
I would have assumed it was fairly obvious, but in case it's not: a million-line codebase (also known as a "harness"), running at inference time, orchestrating thousands of calls to a neural network for any given task, is the exact definition of a "neurosymbolic architecture"
Daniel Lin retweeted
We ran Kimi K3 through 3 agent harnesses (Claude Code, Hermes, Kimi Code) on 28 identical tasks.
All 3 harnesses completed the tasks at similar success rates, but the interesting story is token efficiency: the same task cost up to 30x more tokens depending on the harness. 🧵🧵
Daniel Lin retweeted
@hellominds_ @Ethoswarm growth has been explosive and is one of the fastest growing AI agents platform now number 1. in Entertainent, Roleplay, Creative Writing & overall Top 10 global AI Agent according to @OpenRouter and you can get AI credits using @Mocaverse_xyz $MOCA
ALT Fastest growing personal AI Agent Platform
Daniel Lin retweeted
1/ I just purchased 22 @Mocaverse_xyz @Moca_Network NFTs and will also include 3 ethereum:native for a special giveaway I am running involving @hellominds_ @Ethoswarm in order to compete you need to join an ai minds based competition I created, some (but not all) of the details below 👇
Daniel Lin retweeted
What does the agentic economy need to become real?
We are excited to announce that @hellominds_, our persistent AI agent platform, has completed the live pilot of a new agent-led commerce experience with @Visa.
This milestone empowers Minds AI agents to seamlessly find and buy on a user’s behalf. Underpinned by Visa Intelligent Commerce’s trusted security framework, the agent can act within the permissions the user sets, bringing together personalization, security, and control in every transaction.
On a single instruction, a Minds AI agent can handle discovery, decision-making, and end-to-end payment in one fluid motion.
What’s more? It dynamically surfaces relevant Visa cardholder rewards & benefits directory that helps AI agents to identify and surface relevant cardholder rewards & benefits.
These experiences are based on skills, catch them as they make their debut at LEAP East.
See full announcement: animocabrands.com/announceme…
Daniel Lin retweeted
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
