Jason Brown

The more I use LLMs, the more I'm convinced that more agents is not the solution. Article this week: How to avoid falling into the agent trap. jasonunlocked.com/blogs/articles… What are your thoughts on this? #Agents #AI #Operations

Result? Faster deployment, real governance, and many teams seeing 20-30% productivity gains without the complexity and risk. Full article here: jasonunlocked.com/blogs/articles… What are you choosing in 2026, chasing agent swarms or keeping it discrete + fully traceable? Drop your take 👇

The smarter (and simpler) play: 1) Discrete, focused LLMs for specific tasks (supplier quotes, downtime summaries, compliance checks). 2) Pair them with basic automation (UiPath, Zapier, etc.). 3) The secret sauce: cryptographic hash logging straight into Excel Every output gets a unique ID + timestamp. Full audit trail in one place. Stupidly simple, incredibly powerful.

The more I work with LLMs in real manufacturing and mobility ops, the more convinced I become: More agents is NOT the solution. Most people are falling straight into the Agentic AI Trap. Here’s what smart executives are actually doing in 2026 instead:

This is a great post. What's ironic too is that they've discovered that people can game this algorithmic system with AI, bots, spam accounts, etc. and then they try to punish the people taking advantage of the system that they set up instead of tackling the real problems. But, just to prove I'm not only negative there is a solution I can offer. Go to the middle ground. Content creators should be allowed to specify what 'category' their content falls in and a certain amount of their reach should be guaranteed to be shown to those that mark they are interested in that category and a certain amount should go to your followers guaranteed. That way you know that there is at least some small chance that people actually see your content and have a chance of making a connection. So essentially then boosting or ads increases the amount of guaranteed traffic to your intended audience. People get more meaningful content served, reach improves, ads structure is left intact, and algo can help diversify feeds so it's not a total echo chamber.

Longtime users are frustrated with X because it no longer works the way it used to. Back in the early days of Twitter, your reach was simple and direct. If you had 5,000 or 10,000,20,000 followers, your posts were delivered to them in real time. It was a true chronological feed—your tweets showed up in the order you posted them, and your audience actually saw your content. The “old” Twitter (pre-2016) was built around that real-time experience. Your voice reached the people who chose to follow you, without interference. Today, that’s no longer the case. X relies on an algorithm-driven, engagement-based feed, meaning even your own followers may never see your posts unless the system decides to prioritize them. That shift—from a guaranteed audience to an unpredictable algorithm—is why so many longtime users feel frustrated and disconnected.

@noahkagan LinkedIn is what people will say to your face, X is what people will say behind your back... except it's to your face also. I do agree though that it can get very negative, you've got to know that coming in.

Hot take - LinkedIn is for nice people. Not saying that's always best but the community here shocks me with how critical they are of... everything. Garry Tan makes some skills and instead of talking about it, people just bash it. Company launches some CMO AI thing - all posts I saw were shitting all over it. Meanwhile on LinkedIn the comments and replies are all surprisingly encouraging.

I feel very strongly about this the more I use LLMs.

Because that's how the software companies make more money. Agents weren't created to help the consumer, they were created because they're a nice neat bundle of software that companies can sell. With a technology like LLMs, how do you sell it when it can do everything? Package it into 'agents'. It's one of the most clever marketing ploys of this century.

I’m surely being stupid. But if AI is rather unconstrained by expertise or capacity or to some extent speed Why do we need to divide tasks or departments to 9 agents ( the marketing agent, the optimization agent etc ) to each do one thing. And then another agent to manage the swarm. Cant one agent just be doing it all you know. It seems very skeuomorphic. Will we have HR agents to make sure the agent agents are being looked after ? A office canteen manager agent to feed the agents ? Seems daft

It's largely because AI can do "everything" and "anything". it sees all options as available. Whereas a human has a specific purpose and motivation that create intent and action. It can't be replaced by algorithms that only generate output tokens based on input tokens and training to know what's "right" and what's "wrong".

AI completely fails at what matters most in business: Original ideas, imagination, and creativity. Ask it to analyze your strategy and to identify ways you can better monetize your business, or to introduce a new angle that creates value - and it's completely lost. All you get are generic ideas you've already thought of, or suggestions that make little sense. Yes you can use AI to do research faster and put together a spreadsheet more quickly - but who cares. Those things save you a few bucks over hiring a low-cost virtual assistant, but not a game-changers. Yes it saves you time, and yes you should learn the tools, but there is nothing extraordinary here. There is no, "Wow, I never thought of that!" moment. Until that happens, the overhype is real.

@Austen If anyone is interested, I wrote all about this in my book: amazon.com/Business-Opera…

@Austen It's because their processes and IT orgs are badly aligned and underfunded. A new approach to IT is needed.

One thing that's interesting in talking to big companies is a lot of them are locked out of AI because their procurement process is too slow. * Nobody can try any new tools = no bottoms-up growth * Sales cycles move slowly enough by the time they're buying AI it's old

@nikitabier I see em dashes...

The fortress we are building—and the layers of redundancy—to protect the platform against the AI Slopacalypse will seem obvious in a few months. Whether we use every tool in our toolkit is TBD, but it would be negligent to not have them ready.

@chamath My god, when will people wake up that the answer to broken processes and failed software implementations is not more software.

Today I’m very excited to announce a global partnership between 8090 and EY. EY will adopt 8090’s Software Factory and use it to help their customers break free from slow, costly and failure-prone legacy enterprise software using our AI-native software factory that reimagines the software development lifecycle. EY is a massive global organization with more than 400,000 employees and tens of thousands of customers in every sector of the global economy. 8090’s Software Factory is the new way organizations can move to a focus on building software that is powerfully bespoke, hi quality, easy to maintain, easy to migrate and always consistent and up to date. No drift, no cruft, no waste. Companies that build with Software Factory grow faster, are more profitable and are more adaptable in moments of change like we are witnessing today. Let’s rewrite all the enterprise software in the world. EY and 8090 will work together to do its part.

@elonmusk @tslaming The reason you need a chip fab operation is to subsidize and protect the one off chips going in the small number of space ships, at least until they are mass produced.

@tslaming Starlink V3 has three separate chip designs by SpaceX that are very special

Read a post today that reminded me of one of my all time favorite quotes that applies to a lot of things, but especially business and entrepreneurship. It's from the basketball player Nikola Jokic who said: "If you want to be a success, you need a couple years. You need to be bad, then you need to be good. Then when you're good, you need to fail. And then when you fail, you're going to figure it out. Experience is something that is not what happened to you, it’s what you’re going to do with what happened to you."

@AmeeStelloAI Just so and something to be aware of when using LLMs for technical projects.

@JasonUnlocked Building plans with explicit downside scenarios and stress tests helps account for risks that sit outside typical assumptions.

One big issue that I see with AI LLMs is that they tend to treat highly unlikely or highly improbable events as being almost impossible and do not adequately take them into account when giving detailed plans, such as business plans or code. And, we know that these black swan or long tail events can make or break success.

@buccocapital There is no value in exchanging tokens back and forth without another revenue stream attached such as advertising or subscription based. General search queries don't make sense for these companies if they can't advertise effectively or get people to pay for it.

So Anthropic and OpenAI are just going to…give the consumer market to Google?