At least when I write an autistic info dump people know I wrote it. Why give your voice over to a corpo slop factory?

Heck, I use LLM assistance for coding and I’ve even coded up whole features with the clankers, but giving it the right to speak for me is too much.

I should also add that I read and understand every line of clanker output that I publish for others, so I’m not a vibe coder either, just adhd.

This isn’t a job risk, those doctors will be imprisoned for a very long time, or worse, if they perform anything resembling an abortion. The voters of Alabama wanted that, apparently.

Yeah thats part of being in a job that has mortality as a part of it

If you’re in the critical path of someone else living or dying, you need to be prepared to take the same risks

I’ve done it myself and was willing and enthusiastic about putting my life at risk for something I thought at the time was worth dying for.

If you’re not willing to die to save/help someone else then you’re really just committed to your own survival which imo is just narcissism

The LLM is also training on or reading your emails; my wife was emailing a client and it produced absolute garbage and in that garbage was information the clanker shouldn’t have known unless it read the other emails. That’s probably not a surprise but the implications are staggering.

The opposite end of that spectrum is everyone here who thinks we should just ignore nature and where we came from since we’ve conquered it, or something. I’m not sure what you all are arguing here. We’re still a part of nature, we’ll discover that quickly if society collapses.

> Engineering should not be subjective.

I think there’s a lot of ways to do engineering right. It’s very subjective in many ways. Every project has multiple possible paths to success.

You can build an MCP client really really easy today with libraries and it integrates with inference calls very well. If you want to give a cloud agent access to another api, you need to bills a connector or give it shell access or some sort of sandbox with tools to access that stuff. That’s fine, but how do you then give it auth access? How do you ensure security boundaries? MCP builds those in and the shape (prompts, resources, tools, etc) are good for agentic work.

It’s like asking why we needed browsers when we had BBSes; they serve a different but similar purpose and are build on different abstraction levels.

I like MCP, but most servers and harnesses just dump all of the tools into context each message. Also the default tutorials for everything around MCP say to dump tools this way.

It’s really hard to determine which tools an agent will need on the fly. The best idea I’ve seen was to do a “search tools” tool which an agent would use to find relevant tools, then have an “execute tool” tool which did what it says.

Alternatively, if you’re using an agent that’s good with code, give it something like “port of context” (pctx) which translates MCP tools into a JavaScript package so the agent can chain MCP tools together with code, and they can filter the data down to just what they need. I haven’t used this method much yet but soon I need to improve the MCP gateway in our cloud agent builder product so I’m going to try this one out soon myself.

Agreed 100%. It's sad when the harness floods the context with a whole MCP response and then see the model dumping everything to disk to process it with a script. That's something the harness should do whenever the result is large: dump the json into a file and just say to the model: "your data is in /tmp/foo.json, it's very large so be careful with it", but I don't remember having ever seen that; maybe some harnesses do it? Depending on what the model is after it might want to filter for this or that, no need to load everything in the context.

That’s a good idea too, I might steal it.

You realize that not every user of agents uses them like Claude or Codex on your local CLI right? MCP is the standard for cloud agents. How do you get a cloud agent working in an ephemeral container access to skills? The answer is MCP.

A 'skill' is generic concept - as short set of right-sized instructions for a given cli or api call, it can be applied in any context.

If MCP did not exist today, we wold not invent it.

We'd probably hormonize in basic conventions around json calls, and not much more.

The rest would just be api use / instructions.

LLMS to day are exceedingly good at calling RESTful APIs, the MCP standard provides little advantage.

The advantage of 'skills' is that they are more generic - an Enterprise LLM can evoke 'capabilities' which may or may not involved rpc type calls, and if they do, there will be varying level of instructions provided.

There's almost not point to MCP.

Yea so your answer is to build something that’s like MCP basically. You’d standardize conventions around json, great, now standardize auth. Oauth is nice right? That’s MCP. MCP is literally a restful API using JSON with OAuth.

You’re arguing against MCP but have nothing to offer that isn’t nearly the same thing.

Agreed, not sure what people arguing against MCP are even arguing against. The only valid critique of MCP is that you think the RPC protocol isn't ideal, sure, you could argue about the protocol design, for example I wish there was better support for stateless calls. But why wouldn't you want a protocol for back-end API calls? Otherwise you need custom clients for each possible backend you want to invoke.

Don’t worry he probably got an agent to do it. Maybe ten.

If you participate in this competition you’re effectively giving free research to a weapons company who will use it to improve their killer drones. I’m not judging but that’s what this is. Please correct me if I’m wrong, but I don’t think I am.

I only say this because technologists sometimes don’t think about the reasons behind things, they just see a fun problem to solve.

Relevant blog post someone threw last time something like this got posted: https://calebhearth.com/dont-get-distracted

I appreciate it, I'd not seen this before, great perspective

So I got a chance to watch this fully through, and fun fact, I was in Iraq as an Armored Cavalry officer when this technology was deployed with the embedded Seal team and special forces unit we supported. I'm very sorry this engineer clearly regrets being what he regards as tricked into developing the technology, but from my experiences, his guilt is misplaced. That technology saved a ton of lives, and allowed us to track and apprehend some really bad dudes. I was the operations center battle captain at the time, so I had a really high degree of knowledge of the information being used to decide on targeting, and it was 100% professional. Huge dossiers on mass pitt style murders, with video evidence, that was required before a target was actioned on, after months of planning. I have zero regrets; this developer should be very proud of his work. I acknowledge that in theory some bad actors can abuse technology for ill, and this current admin definitely disappoints me constantly. But I'm glad we had this tech when we needed it, so as many of my friends could come home as did. We had over a thousand casualties in Iraq over my two deployments, and my unit was 100% all about helping pursue stability and local governance, no nefarious story telling I ever saw. Was it worth all of those lives and money? I'd say no, but only because we as Americans and our system of government doesn't allow for long term thinking, planning, and persistence required to make generational change happen.

Very relevant thank you

Anduril is far ahead of anything we can produce here for this competition, but it's certainly a great method for them to identify potential talent who can bring needed skills to their team, hence the grand prize of a job offer. The terms of the competition is they get to use any submission for marketing purposes but the IP remains with the participants. Personally I'm in the camp that the world has bad guys, and they are building technology too. Smartest to make sure yours is better. Professionally we only work on defensive systems, which are in high demand these days.

We can defend from drones without building our own autonomous drone killing machines. The real danger here is that we have bad guys at home who will use this tech on us if we give them the chance, so personally I’m opposed to building it. Not like I can do anything other than comment here and vote.

I’m really not judging anyone who does, there’s a lot of nuance here that honestly makes it hard to argue you’re wrong or I’m right in this discussion.

Completely agree, details matter for sure. I just returned from a visit to Odessa Ukraine, and observed a maelstrom of drone on drone combat right outside my hotel window to booms and shaking glass. Those experiences make me want to help defend against the indiscriminate strikes however I can. But I concede there is danger in providing the tools.

> Personally I'm in the camp that the world has bad guys, and they are building technology too. Smartest to make sure yours is better.

I get the idea, but it is also the same argument that has fueled the nuclear arms race that stuck all of us in this permanent launch-on-warning, mutually assured destruction situation.

As push comes to shove, I'd prefer our drones are better than Iran/Russia's.

The only reason we’re in this mess with Iran is a dumbass president; as for Russia, same thing. I’m not gonna reward the military industrial complex just because they backed a moron.

On topic: I would prefer we research defenses for drones instead of drones themselves.

I would prefer our drone defenses actually work against real drones, and that only happens by having real opponents. Or do you not believe in red teaming?

That isn’t real opponents, it’s useful for training but you don’t need actual drones to red team. I don’t know how many times I had to use my imagination in the military when we were fighting paper opponents that were non existent but we had to maneuver and walk through processes as if we had a missile inbound. Or those situations where we had a real target but we weren’t allowed to actually shoot at it because it was reserved for another unit to actually shoot at.

Red teaming is fine but we already have the actual red team drones to make sure they work against.

I totally agree, very interested anti-drone defenses. You can actually use much of this to model and simulate the drones you want to stop just as well.