Christian Labon

CEO @UltimateAuto_

[Christian Labon @christianlabon]

@SaidAitmbarek The problem isn’t the agent trust, it’s the knowledge about scoped profiles and utilizing lightweight local models to parse and sanitize sensitive data before it touches a cloud model that makes this possible. Or have scripts do this automatically…

[Christian Labon @christianlabon]

Pay for the hardware and the model that best serves you, recreate the tools yourself and keep them forever. Too many companies capitalizing on basic tools that your agent could build suited to you in an afternoon.

Then link those tools together and watch the magic happen. A singular AI based tool becomes the framework you transform your company with extremely fast.

[Christian Labon @christianlabon]

I guess that depends on if you’re a Dev or someone implementing Ai into an industry that hasn’t adopted it yet. For me AI has greatly increased the number of hours I work by choice, because the payoff in the long run is worth the effort.

Even in the short term automating the monotonous tasks and giving me the time back to focus on my core business and designing AI around that business is a game changer.

[Christian Labon @christianlabon]

@Enscion25 Trying to diversify and appeal to mass consumers way too early. The technology is frontier, the interface surfaces are not. Hoping they get back on track because once you figure out how to properly harness Astra the results are incredible.

[Christian Labon @christianlabon]

Doesn’t take much to open the mental flood gates of ideas. Pick a software or hardware idea you’ve had in the past that you didn’t have the knowledge or time to do yourself. Have your agent execute that task, all of the sudden all of the random software based projects you’ve ever had will surface and you’ll be burning tokens in no time. Homelab, 3D printing, Personal assistant, Inventory tracking, connecting services that don’t currently integrate with each other, the possibilities are endless.

[Christian Labon @christianlabon]

The bottle neck is also the orchestrator. Astra is a model so capable and focused on frontier level intelligence that it would rather go find and engineer solutions to edge cases than follow a natural language prompt. Use Opus 5.5 as the orchestrator and experience the true capability of Astra.

[Christian Labon @christianlabon]

@johnHarry1018 As soon as humans get involved in the loop and stop posting what their Agent says will do the highest numbers on X we will start to see the shift. The knowledge is already out there, just have to pick up the crumbs as you find them.

[Christian Labon @christianlabon]

@TTrimoreau If Anthropic can manage to keep the blinders on and continue to ignore the AI fads that everyone else follows and poorly copies every week and continues to just develop badass models with fair usage they may hit a pace nobody else can catch…

[Christian Labon @christianlabon]

@NousResearch @tomfgoodwin The consumers that learned about the power of muse off of instagram and tik tok with no knowledge of the risks that come with the power of AI are going to have to learn these lessons the hard way. Just hope it doesn’t burn people so bad that they write off Agentic AI for good.

[Christian Labon @christianlabon]

In my opinion it’s just trading off the short term pain for long term gain. Once you shift from basic usage to any sort of “power user,” the lack of control, oversight, and security ends up costing you the same amount of time moving everything over to a harness like Hermes.

Grok bot and the other consumer platforms are a great way to get your feet wet, but use them to set up an Hermes agent based on how you use it and you’re a lot better off in the long run.

Do yourself a favor and start a local vault with all of your .MD Files so that when you inevitably switch over you don’t have to start from scratch. Allows you to be model AND harness agnostic.

[Christian Labon @christianlabon]

@nikpolale Yes, but not as a main orchestrator. Opus 5.5 runs the show, 6.1 Sol or Astra do the work. Will completely transform your opinion on Open AI models when you let a model as powerful as opus 5.5 do the prompting.

[Christian Labon @christianlabon]

@tomosman I really hope we see another startup give us this feature for the 15th time this month, the last 14 really didn’t do it for me.

[Christian Labon @christianlabon]

@Teknium so crazy that you have to comment on the fact that you are not going to follow the trends and build the same rushed and extremely beta slop that most of the other consumer focused companies are slapping their name on.

The core behind Hermes is the magic and the refusal to wrap it’s use cases up into a small box of gimmicks like everyone else is exactly why I don’t get distracted by the all of the shiny new objects shipping every week.

Don’t be mad @nousresearch won’t do the work for you, point your agent at the feature you like and build your own version tailored completely around you and your specific use case. And use that feature for as long as YOU want to use it, instead of using it until the company that shipped it decides to swap it out for the next shiny object and you end up right back where you started.

Too many people see the power of Agentic AI and assume that means they don’t have to put the hours in to get the results. Stay the course, ignore the fads, keep building for your core user base! The industry is moving too fast and the technology is too far from a set it and forget it mentality to be trying to tailor to the consumers that have never used AI before. Either way the ones who truly take AI seriously will make their way to Hermes in the end once they realize everyone else is only building products to stay at the top of the tik tok news cycle…

[Christian Labon @christianlabon]

If only they could focus on building the best models instead of porting the best products. They should want all of the major players to run their inference on Open AI instead of trying to be the jack of all trades in the consumer space.

Grok bot running on Opus 5.5 and routing jobs based on best models shows just how much potential simply making the best model can have.

OpenAI is the jack of all trades master of none, anthropic is giving a masterclass on being a master of one. Not getting distracted by the shiny objects of image gen models, early beta consumer products, and copying every new startup that gains real traction and wrapping it up into their own ecosystem in a rushed and half hazard attempt to claim the top spot before the startup can claim authority in the niche.

[Christian Labon @christianlabon]

Exactly what I’ve been preaching. Astra does not respond well to natural human language and input and does not how to keep itself from going down never ending edge case rabbit holes. Opus 5.5 is the brain that lays out the blueprint, Astra is the team that builds and executes. Opus comes back when it’s done to make sure it’s all up to code

[Christian Labon @christianlabon]

Absolutely agree, but that falls more on lack of experienced, discretion, and proper testing internally before pushing to production. But considering for many people this is all new territory it’s easy to see how the excitement about turning your ideas into shareable reality can overshadow the fundamentals that they may or may not know to follow.

[Christian Labon @christianlabon]

@Perplexity_AI normalized the concept, unfortunately they lost the steam needed to stay relevant to most consumers. Perplexity has been first to deliver to consumers on so many of the things we put value to in the space today. The day perplexity personal computer launched was the day my eyes opened to the potential of agentic AI. Credit where credit is due!

[Christian Labon @christianlabon]

Can I lock in my vote for a reset on day 17 now then? In all seriousness love all of the effort on improving the experience for current active users, would just like to see the updates and changes being asked for the most. Would take 5 days of the most requested updates over 28 of the ones that are largely inconsequential.

6.1 sol ultrafast day 4?? 🙏

[Christian Labon @christianlabon]

You can’t get randomly banned by a backend error and lose all of your data for another…

Too much time and value in what my agents do and the system I’ve built to rely on a 3rd party cloud to keep it safe. That along with the ability to run jobs on local models to parse and sanitize my data before giving it to a cloud model to make use of the data is a workflow that is not discussed nearly enough.