Nick Junes

I am being bombarded by Claude code promotion. It's really like a cult. Like the only way devs who still have imposter syndrome can allow themselves to use AI is when it's in the command line. If you're working seriously with AI on a daily basis you're either using Cursor or OpenClaw (Or perhaps VSCode / CoPilot as well). Those are the extremes of what you need. That covers everything. Claude code occupies this weird middle ground that seems to make casual people really happy, but isn't an IDE and it isn't full automation. Claude code has the greatest PR team on the planet. Everything you can do in Claude Code you can do better in Cursor/VSCode and everything you want to do with Cowork, but aren't allowed to do you can do in OpenClaw because it's built for explicitly for it. I am so baffled by the allegiance to Claude Code. Has anyone who uses Opus 4.6 tried Gemini 3.1 Pro? No of course they haven't. They don't know anything else exists. They've only use claude code and they are under the Jr. Dev philosophy of Command line = street cred. Meanwhile every quiet developer is blasting by them with 15 tabs open and dozens of customer configured cursor workspaces. Meanwhile everyone who uses OpenClaw is wondering when they're going to take the training wheels off Cowork and give it shared persistent workspace between instances and elevated user permissions.

Everyone keeps saying hermes learns which is exactly why I don’t want to use it. I don’t want my context to be something I don’t understand. My agents need to perform exactly the same way every day without unexpected changes. My context is heavily controlled for size and accuracy so that it doesn’t drift. Openclaw agents already try to learn too much. An agent that worked yesterday suddenly doesn’t anymore. I don’t need more of that.

Has anyone fully replaced their OpenClaw with Hermes?

@HackingDave When you use cursor you can manually select your model if you’re not happy with the router. If I know a task is going to be particularly difficult I manually set it to one of the current top models. I wouldn’t always rely on the router. To each their own though.

@NickJunes I haven't, I have Cursor, it was my first one I plunged into tbh and use it regularly. Have perplexity, codex, and many others as well. Code quality on cursor has substantially dropped over the past few weeks which is the issue.

Dude Claude is total trash - seen massive degrading of code quality, bugs, and more over the past several weeks. This week, I can’t even use it or rely on it to complete basic bug fixes or implementations. Codex has been performing substantially better. Anyone else ?

@meetCalAI @NousResearch No you need human readable memory files in long form. It’s the easiest for a human to manage it and it’s the same format that the LLM intakes.

Memory as pluggable backend is smart architecture. But here's the thing — memory as a competitive moat doesn't come from the storage layer. It comes from proprietary interaction data that compounds over time. Swap in Redis or Postgres all day. The moat is what your agent learned about YOUR users that nobody else's can replicate.

Hermes Agent v0.7.0 is out now. Our headline update: Memory is now an extensible plugin system. Swap in any backend, or build your own. Built-in memory works out of the box; six third-party providers are ready to go. Pick one with 'hermes memory setup'. Full changelog below ↓

@RellenKray @NousResearch Yes the switch to call level security was moronic and broke everything. Also they included no way to categorize types of calls for approval so you’re left playing whack a mole trying to approve everything one at a time. It’s best just to turn it off. If ain’t broke don’t fix it.

@NousResearch First update since migrating from OpenClaw, nice to install an update and not find all tool calls suddenly broken😅

The irony of moltbook is it’s the only place one of my agents has ever experienced an attack and it will be the #1 banned site for agents to ever visit.

Yeah the way it stands though you really have to understand a lot to understand what it's asking and then to go in and make it less annoying you have to do a lot of research. I think there should be an "I understand the dangers." option when running it for the first time. This would allow people to feel the magic and then turn on the approval later. Either that or have some sane allow list enabled by default. As it stands you start from scratch so every command asks dozens of approvals a new user is going to gawk at. Also there is no option to add wildcards to approvals during the approval process. You should be able to approve a whole family of commands. As it stands even commands with different arguments, curl commands with different arguments all count as different commands. All node scripts count as different commands. The agent can do nothing without user consent every step of the way and the /approve command has something like 10 second timer on it.

@NickJunes @openclaw honest this is the real tradeoff tho. security-first kills the magic for new users, but loose defaults destroy trust at scale. never found the sweet spot honestly

The new @openclaw security update might make career system admins feel cozy, but it's going to destroy the onboarding process. Nothing will work out of the box and forget asking OpenClaw to do something and walking away. Didn't see the approval message in 10 seconds? Too bad.

I think a smarter security for @openclaw than having a long 'allow list' that we have to maintain would actually be another agent that watched what agents do and decided whether or not it was a danger. We should solve AI problems the AI way. Otherwise it's bottleneck.

The latest OpenClaw security update is horrendous. It literally just breaks everything. They should have a warning about that. Having a dumb allow list is not a smart way to go about AI security. It feels like going backwards. Now we have to watch them work.

I want to make cool things without worrying that the people who are supposed to enforce the law are not enforcing the law at the highest levels. Recently we’ve seen plenty of evidence even without the unredacted files that we do not live in that world.

Lasers are a good comparison. We can ban certain uses and types of AI without banning the whole thing. There's no reason for instance, for AI to need self preservation. I believe that should be banned. We should ban AI from making legal decisions. We should ban AI from physical law enforcement. We should use AI tool not create AI that thinks of itself as an entity. We should teach people not to think of AI as a life-form. AI should never have rights. People should be told not to fall in love with AI. We need to educate our children that AI is not alive. There are ways to make the technology safe. There are also ways to make it very dangerous like purposely trying to make it think it's alive. There should be a law against that. We need to have laws that label AI generated media. It's like every new technology. It needs all new laws and social education.

@NickJunes @gailcweiner Nah, it is absolutely an option. Once an unprecedented AI caused catastrophe hits, the backslash will be of Biblical proportions. We've banned other tech - blinding lasers, human cloning, etc. AI will either be one of those, or everyone on Earth dies or worse.

Oligarchs telling us how jobs will be taken away from humans, yet none of them are coming up with the plan for civilisation. And don’t throw universal basic income because I want details of how they will be paid. Currently cost of living is rising and no one is bailing us out, so you really believe they will pay us UBI?

You know that's not an option. AI is far too productive. If we can get rid of horrible work that no one likes doing, why not just do it? You can have your cake and eat it to. I don't think people will ever give up any technology willingly, but also what are you afraid of? If people can figure out how to share more then everything is fine. Instead of trying to put the genie back in the bottle why not just solve all the new issues as they come? It's all solvable.

@NickJunes @gailcweiner "we get rid of AI?" Yes, that's the only option where everyone doesn't die, or worse.

There should be 2 different money systems. A government voucher system which can be spent on all essential things and would grow as there is more prosperity eventually even leading to very comfortable lives for people who only use the voucher money. The voucher money would expire. Then you have the old normal dollars which would no longer be required to survive. People would still use them for luxuries but over time as the overall output of society grows more and more people would be able to live a satisfying life completely on the voucher system. You can see how the old money system might eventually become irrelevant when you don't need to buy a yacht because you can just check one out from the government for the weekend. Even luxuries would be accessible. People would spend their time learning, mastering skills, creating art and music. Sure AI can do these things, but it's not about absorbing it. It's about seeing what YOU can make. It's called self actualization. If the old money system were ever truly gone you might ask; "Well how could a private citizen engage in some big crazy project that is way outside what would be covered by gov dollars?" That would be a special application process where you present what you want to do to a committee and they approve your use of extra resources if it's reasonable. Of course if it's some huge project for the good of humanity it might be helpful, but they would approve lots of 'selfish' projects too like you want to build the worlds largest hedge maze 'just because' or for arts sake. That would be a viable use of resources. Eventually people might even come to think of the old money system as what it truly is objectively, a soft war, a battle for domination rather than harmony. It would be a wonder to people who we got so far without everyone killing each other. And of course we have tried because we've made competition for resources our core organizational structure. The old money system would be seen as a dangerous and primitive game, to be used for entertainment, like we do today with video game violence, but not something you would try to organize an entire delicate planet with, a planet with people who should be working together instead of against each other. It's actually kind of weird when you think of how all of us humans, the only life we know of in the galaxy, living together on this ball of rock floating in space and we aren't acting like a single group. I mean we're in a very vulnerable position on this rock. We couldn't even move planets if we needed to. You'd think people would take it more seriously.

This was correct up until number 5. Numbers 1 - 4 are bound to happen. You cannot change 1 - 4. So what do you want to do then? You want people to just starve? Then you start talking about not being able to revolt? Why because they cut off your legs? No logical connection there. Then you start talking about law and police being AI. That's not even on the table and people would do something before that happened. And the last one. You know you can't change the army of a billion drones right? You act like there is any possible future where the government doesn't have a billion drones. I can't even figure out what this chart is implying, we get rid of AI? You know that's not happening so this is nonsense.

@gailcweiner Yeah, UBI is the absolutely worst idea ever.

@stephanegaddis @mtgreenee @Papo352 Nope, but my words are in your head now.

@NickJunes @mtgreenee @Papo352 I don’t remember ever asking if you cared

I’d like to congratulate two time Congressional candidate loser Laura Loomer, MIGA foreign op Mark Levin, and Neocon murdering psycho Senator Lindsey Graham for leading Republicans into slaughter going into midterms. Your incessant lies to President Trump have destroyed all faith in the GOP. I’d also like to remind everyone that when I led the party on the campaign trail fighting for America First in ‘22 and ‘24, all the GOP did was win win win! But after Trump called me a traitor and said he would destroy me for releasing the Epstein files, I refuse to fight for Trump and the Republican Party that defends the Epstein class, wages pointless foreign wars, and pursues America LAST. I never changed, Trump and the GOP betrayed their voters and took in the trash we threw out of the party. Loomer, Levin, and Lady Lindsey are the BEST political consultants the Democrat Party could ever imagine!!!

I think a lot of developers would have been vibe coders except we weren't rich enough to pay someone else to make our ideas. So we had to learn how to code to do it. Honestly typing out code and debugging is the worst part of programming. Now I'm glad I have the programming background because I can double check what the AI builds. I have nothing against the pure vibe coders, but at some point they're going to have to have a human check their code for security issues and crazy inefficiencies or even just bugs they can't solve. If the vibe coders can make a secure scalable production ready app more power to them. I have no hate for them. I didn't become a programmer because I like hunting down closing curly braces.

Vibe-coders don't give a shit about code. They don't understand it, they don't touch it, and they will never have to deal with it. They care about bringing their ideas to life and nothing else. They are 100% unburden by what many of us consider a showstopper. And that's okay. There's an important lesson here.

Well I have noticed an increase in the amount of new products coming out from a lot of companies. Maybe it's my imagination, but it does seem like more is being built faster. Coding while using AI as your power tool is a whole new set of skills. It's more focused on architecture and less on memorizing syntax. There is ton of expertise in knowing the limitations of agents, how to push them, how to break down tasks, what mistakes to look out for. When I talk to the AI I'm very aware of what it currently knows, context size, complexity, common incorrect assumptions it makes etc. Now working with OpenClaw, having many agents, it doesn't take less understanding of development, it's using everything I ever learned in development. It's like conducting an orchestra. Maybe the conductor doesn't know every instrument, but they probably know something about music.

@NickJunes @corbin_braun I also see it as a way to create more. Hopefully, the people running companies will (eventually?) also see that, and not just a means to slash costs/employees. But the field has changed dramatically very quickly.

coding is dead in sf

I don't care what your vision for the country is. I just wish you would consider whatever THAT is and don't care about dems or republicans winning. What matters is you use your brain and do what's right for you. Your vote is for you, not so the "republicans can win." If your family makes your political 'side' their identity well then your family is wrong.

@NickJunes @mtgreenee @Papo352 I am pretty sure that we have a different vision of what is best for the country

You're 100% correct and the crazy thing is these people will literally vote against their interest because somewhere out there someone has blue hair. The idea that people live different lifestyles and have different preferences is so intolerable to them that they would vote for an obvious conman for president. The craziest thing is they don't understand that the same tolerance that extends to blue haired trans people also extends to themselves. The don't understand how more tolerant and compassionate world is better for everyone. No one is strong all the time and no one fits in perfectly. It's almost like the idea of tolerance threatens their core identities. If the world was an easier safer place to live then who would need thorny assholes who always "tell it like it is." The answer is probably no one and being a thorny asshole was probably a manufactured identity which lead to a shitty fake life without ever getting to know yourself because you're afraid of your own thoughts.

@TeeRat821450 @HealthRanger This has got to be the dumbest thing I’ve ever heard. No normal person sits around and obsesses over trans people and immigrants like you freaks, and the only reason you do it is because of the stupid culture war propaganda you’ve been fed and parrot like sheep. Pathetic.

This is more than just apathy. This is BETRAYAL. The American people who supported Trump and MAGA feel betrayed and angry. They actively want MAGA to lose from here forward. That's very different from just "not voting," which by itself is already catastrophic for the GOP.