paulgb

Neat, but what’s to stop a server from reporting one innocuous set of tools to MCP-Shield and then a different set of tools to the client?

Manhattan recently got congestion pricing and it’s noticeably more enjoyable because of this, even indoors.

Is congress actually apprised of the tariff developments, or do they learn about them from twitter like the rest of us?

This isn’t really “allowing” fraud though. From the memo:

> Prosecutors shall prioritize cases that hold accountable individuals who (a) cause financial harm to digital asset investors and consumers; and/or (b) use digital assets in furtherance of other criminal conduct, such as fentanyl trafficking, terrorism, cartels, organized crime, and human trafficking and smuggling. Seeking accountability from individuals who perpetrate these types of wrongdoing deters future illegal activity, compensates victims, and promotes the public’s confidence in the digital asset markets and broader industry. On the other hand, criminal matters premised on regulatory violations resulting from diffuse decisions made at lower levels of digital asset companies often fail to advance the priorities of the Department.

I read this as switching from proactive enforcement (stop the act before it causes damage) to reactive regulation (punish the guilty after the act happens). Proactive regulation of public markets makes sense because the general public has an interest in well-functioning capital markets even if they are not direct participants (markets are essentially the capitalist replacement for central planning).

I worry that having the SEC regulate crypto can actually be worse for those people, because it sends the message that it’s a legit government-approved investment instead of a wild west zero-sum craps table.

My read is that they will still prosecute fraud once it happens, they just won’t proactively regulate the industry.

I’m not a fan of Trump but my steelman argument for this is: policing public markets is a public good because efficient capital markets are the bedrock of a capitalist society.

It’s hard to make the same case for crypto, which (despite the insistences of its participants) more or less exists in isolation from the rest of the market. The industry feels it is over-regulated. Makes sense that a more lassiez-faire leader would take a more hands-off approach to it.

I think the only reason people think MCP is over-engineered is that people have been waiting for a way to expose remote tool calls in a client application, MCP came along and had that (among other things), and people assumed that that was all it was good for. But it's not! It's like saying a microwave is over-engineered because you only use the timer functionality.

Interesting, I could be wrong then! The inclusion in his story of some very specific irrelevant details (like his cell/block number) while omitting anything that could be verified (like which border crossing and detention facility) was one of the things that raised my suspicion, but if he's sharing specifics now it raises my odds that he's telling the truth.

His story is a bit suspicious to me, I believe that he was detained but I can't help wondering if the rest is fiction based on watching US prison dramas. He would not have been sent to a federal prison for one thing, and his depictions of violence are what I would expect from long-term convicted felons, not people being held for having the wrong visa.

> At least one visitor from the UK got locked up in a facility for a week.

If you're talking about Rebecca Burke, it was almost three weeks. All for something that was at worst a naive understanding of the visa she needed.

https://www.independent.co.uk/travel/news-and-advice/rebecca...

> MCPs are opaque to the user and devs (unless they look at each source-code and pinpoint each inspected version).

This is true, but also generally true of any npm dependency that developers blindly trust.

The main difference with MCP is that it is pitched as a sort of extension mechanism (akin to browser extensions), but without the isolation/sandboxing that browser extensions have, and that even if you do run them in sandboxes there is a risk of prompt injection attacks.

Those companies run ETFs. They “own” lots of companies because they sell broad market funds to basically every passive investor in America.

> the very people who perpetuated suppression of legitimate dissent and honest truth whine now that their free speech rights are being somehow abrogated

Are you implying that Brian Krebs perpetuated suppression of legitimate dissent? Care to link it?

Standards might not have been _the_ problem, but they were _a_ problem. Before MCP I couldn't distribute a tool call that could be configured by end-users to be used with multiple mainstream clients.

That article seems to miss the point by being incurious about _why_ there is hype around MCP instead of LangChain, LangGrah, SmolAgents, LlamaIndex, etc.

We've had tool call frameworks before, but we haven't had a way of making tools that our clients actually talked to. There was no way to build tools that other people could download and run locally without getting them to switch to a client that baked those tools in. It's like the difference between static linking and dynamic linking, but for tool calls.

They have a PEO option, but FWIW they can also be used as a payroll provider / HR system (benefits access, vacation tracking, etc.) without a PEO.

I linked to the press release because I know X links are frowned upon, but I think it doesn’t do the story justice.

Parker Conrad’s X thread is good: https://x.com/parkerconrad/status/1901615179718406276

Here’s the full complaint: https://rippling2.imgix.net/Complaint.pdf

When it came out, block-based editing and always-on wysiwig were novel, and Notion was definitely more delightful than the existing “internal wiki” software category (Confluence etc.)

One thing I like about MCP's decision _not_ to just use HTTP APIs is that it makes it possible to build and distribute tools that people can use without having to set up a server for them. This goes a long way for enabling open-source and locally-running tools.

Transaction fees currently make up a tiny sliver of the mining rewards, and this seems likely to be even more the case the more ways there are to exchange it off chain (like trading ETFs). https://mempool.space/graphs/mining/block-fees-subsidy

> as of now, Bitcoin hasn't been hacked

We should be cautious about whether past precedent applies to future bitcoin, because the bitcoin halving mechanism ensures that bitcoin enters uncharted territory (with regards to the incentives at play) every four years for the next ~100.

Satoshi's design was that bitcoin would be used as cash and transaction volume would provide the funds to secure the network. That largely hasn't happened (it's a "store of value" now, lightning exists, etc.), so mining incentives are still ~95-98% funded from the exponentially diminishing supply of new coins.

Short of some institution stepping in to secure the bitcoin network by mining at a loss, it's hard not to imagine a 51% attack being relatively cheap on the time scale of decades.

I'm holding off on using it until the IDFK license is OSI approved.

In the article, there’s an example where a graph object needs to be:

- converted to png

- converted into a promise of an array buffer

- awaited

- converted into a Uint8Array

- passed to Deno.jupyter.image

Just to display it in the notebook. In Python Jupyter, you can just return many types and the library/Jupyter will figure out how to render them.

A few lines of code isn’t much but in my experience for interactive coding to be productive you have to remove every bit of friction and boilerplate you can.

In NYC it's hard enough to get people not to call them e-bikes (even when they are gas powered), I've given up on the scooter distinction (which then also becomes ambiguous with what I can only call "kick-scooters", most of which are not kick-powered these days).

Only if you give it unfettered accesss. AWS has an API called AssumeRole which can generate short-lived credentials with a specifically scoped set of permissions, which I use instead.

Thanks, I’ll send you an email!

ChatGPT's code interpreter is mostly used as a calculator / graphing calculator. It can run arbitrary Python code, but it is limited in practice because it can't (e.g.) make external web requests or install arbitrary packages.

This is meant to be usable for those use cases, but also to allow apps/agents to make API requests, load data from various sources, etc. It can also run in a company's cloud account, for compliance situations where they are running inference on their cloud account and want a ChatGPT-like code interpreter where data never leaves their VPC.

We allow outgoing requests because a common use case of ForeverVM is making API calls or fetching data files (the "fetch and analyze data" button shows an example of this).

We give every repl its own network namespace and virtual ethernet device. We also apply a set of firewall rules to lock it out from making non-public-internet requests.

Good insight! We also initially tried to use Jupyter as a base but found that it had too much complexity (like the widgets you mention) for what we were trying to do and settled on something closer to a vanilla Python repl. This really simplified a lot.

We've generally prioritized edge case handling based on patterns we see come up in LLM-generated code. A nice thing we've found is that LLM-generated code doesn't usually try to hold network connections or file handles across invocations of the code interpreter, so even though we don't (currently) handle those it tends not to matter. We haven't provided an official list of libraries yet because we are actively working on arbitrary pypi imports which will make our pre-selected list obsolete.

> Love what ForeverVM is doing and the early promise here.

Thank you! Always means a lot from someone who has built in the same area.