Paulina Sáez (@PSaezT) header
Paulina Sáez (@PSaezT) profile picture

Curious, skeptical. https://xitter.gq/t.co/KG6UYcWwWZ Smart business systems

Marbella, Spain
Joined May 2010
Photo by Paulina Sáez (@PSaezT)Photo by Paulina Sáez (@PSaezT)Photo by Paulina Sáez (@PSaezT)Photo by Paulina Sáez (@PSaezT)Photo by Paulina Sáez (@PSaezT)Photo by Paulina Sáez (@PSaezT)Photo by Paulina Sáez (@PSaezT)Photo by Paulina Sáez (@PSaezT)Photo by Paulina Sáez (@PSaezT)Photo by Paulina Sáez (@PSaezT)
You don't need to know how to code to use Claude Code. You just need good management skills.
3
4
343
One of the most important lessons I've learned over the past few months working with AI is how to overcome the curse of knowledge. We tend to forget that when we ask someone to do something, they don't necessarily have the same information we do. Then managers get frustrated when people don't deliver what was expected, when the real problem is that the task was never clearly explained in the first place. Ever asked an AI agent to do something and had it get the task completely wrong? Before blaming the AI, ask yourself: did you actually explain what you wanted clearly enough? AI is one of the best tools for learning how to delegate effectively. And that's a crucial skill for scaling your business.
3
4
184
Something that's worked really well for me is exporting sessions where an AI agent struggled with a task, then using those sessions to create skills or SOPs. In this case, an agent struggled to recreate a DOCX document from a screenshot. Now, even cheaper models produce much better documents using the skills I created from that session.
45
3 things I learned from the Model Thinking course: 1. You'll understand more about a system by looking it through different frameworks. 2. Scenario analysis is one of the best stress tests to find out how the model behaves (and if/when it's useful). 3. A model's job is not to do a failthful representation of the future, but to be good enough to explain and/or predict reality.
Scott Page's Model Thinking course is one of the best courses I've ever taken for understanding how systems work in the real world. His book (The Model Thinker) is a great starting point. But if you can, I'd highly recommend taking the full course.
85
The reason you don't need to know how to code to use AI effectively is that most of it comes down to skills and good habits. Want to manage your context window better? - What an engineer might do: Set up a hook to trigger automated alerts when certain thresholds are reached and optimize subagent usage. - What often works better: Start a new conversation for each topic and only upload relevant reference materials. Good habits will get you 80% of the way there.
3
167
Plenty of people have made good money by buying a business and improving just one area where it was falling short. For example, buy an old-school cleaning company and introduce digital systems to manage jobs and streamline operations. Today, a real competitive advantage is making your business easy for AI agents to work with. Make your forms and customer-facing platforms accessible to AI agents by adding an API or MCP server. (And the real GOATs go one step further and provide a custom prompt to help customers set everything up with their own AI agents)
he comprado una gestoría y la he expuesto por MCP. ahora tu agente puede darte de alta de autónomo, crear tu SL, emitir tus facturas y preparar tus impuestos. un asesor fiscal humano revisa cada declaración, y si nos equivocamos, la multa la pagamos nosotros. try it: @melso_ai
5
3
180
Scott Page's Model Thinking course is one of the best courses I've ever taken for understanding how systems work in the real world. His book (The Model Thinker) is a great starting point. But if you can, I'd highly recommend taking the full course.
I love how everyone is now targeting “systems thinking” as the most valuable skill in tech. Guys, that was always the answer
1
142
Last week I got the flu and was knocked out for the entire week. I finally feel ready to get back to work. LFG
1
39
You can disagree with a lot of what Elon does, but you can't deny that he's worked his ass off to make an impact on the aerospace industry.
great day to rewatch this old video of Elon
53
All my agents start from the same folder. Here I have my main skills, config, and agents.md files. This includes a Wiki where I store my decisions, information about myself, my business, etc. This is extremely useful for doing non-code work.
A lot of folks ask me "how do you work with agents in multiple repos at once?" I have a very simple solution - the Home Repo I have a single repo (the home) where all of my skills, steering files, wiki etc live. I run my agent in the home repo, and then inside AGENTS.md I have pointers to the other repos: ``` For skills, clone mattpocock/skills to ~/repos/skills. For Sandcastle, clone mattpocock/sandcastle to ~/repos/sandcastle ``` This is extremely simple and incredibly effective. The home repo lets you: 1. Put all your steering files and tooling in one place 2. Have a single spot to run all your agents in 3. Make sweeping changes over multiple repos Talked about this in the course yesterday and folks are loving it (see below). I'm sure you will too. Enjoy!
2
101
When I was in high school, I told a career advisor I wanted get into med school. They told me that, since I didn't have top grades, it was unlikely I'd get in. I got accepted into med school two years later. Then I dropped out and went to business school. When I told one of my finance professors I wanted to work in corporate finance, he recommended broadening my options because he thought I was unlikely to get a position. One year later, I was working as an intern in Scotiabank's Corporate Finance department. Then I got hired as an analyst in the same field, but this time at Deloitte. People give advice based on their own experiences. Some things that seem almost impossible to others might be perfectly achievable for you. Just because someone thinks you won't make it doesn't mean you shouldn't try.
“Get fucked” is always an appropriate response to someone who tells you your dreams are too big. They’re too big for them, not too big for you.
1
73
"The hardest part is figuring out what the customer needs so we can build something useful"
the hardest part now of our job is getting the requirements correct
71
I feel Terra 6.1 has been nerfed as well. It's made so many dumb mistakes in the past few days. Sonnet 4.6 keeps rocking tho.
Anthropic nerfed the Sonnet 5.5 plan so bad it's pretty much useless now LOL
1
103
Armored MMA?! TYSM algorithm.
61
I've just finished setting up this email triage automation workflow. For now, the agent will only categorize my emails. I'll run a trial for a week and see how it works.
If you're still checking your emails manually, you have yet to understand the true value of agentic systems.
1
90
Why are models not aware of their own capabilities? Is there a reason for this? ChatGPT voice is even worse. It says yes to everything and then does nothing.
2
1
150
Grok said the risk comes from the power surge after a power cut. We had one yesterday so I'm getting a couple of these for the OLED TV and my MacBook + Mini PC.
Darf man einen OLED Fernseher echt nicht vom Strom trennen? Ich habe das gerade zum ersten mal gelesen. Und was ist dann bei einem Stromausfall? 😳
95
What do you mean "Get ready for Christmas". Halloween is just over.
1
53