The Artificial Investor #59: Personal development in the AI era

 ✍️ The prologue

My name is Aris Xenofontos and I am an investor at Seaya Ventures. This is a special edition of the Artificial Investor about AI-linked personal development.

Special edition: Personal development in the era of AI

AI agents automate manual processes, such as document processing, data classification and analysis. Ask ChatGPT anything about mathematics, physics or biology and you get (mostly) a perfect answer. AI tools automate software programming and data analysis.

In a world where AI is taking over most knowledge worker jobs, particularly the entry-level ones, while institutional education is playing catchup, how does one prepare and make themselves valuable?

Azeem Azhar has talked about the Exponential Gap, the gap between the level of sophistication of technologies that move at an exponential pace compared to the slow flat-line pace of adaptation of society creating an ever-growing gap between tech-savvy population and tech laggards. How do you avoid getting stuck in the Exponential Gap?

Let’s dive in.

We believe the key is in the transition from knowledge-based to skills-based training. Of course some introductory knowledge is important, but the focus should be on skills.

Let’s split AI education in four different areas. Which one you go for will depend on the level you expect to require for your job.

1. Understanding AI and where we are heading
2. Becoming a conversational AI power user
3. Become a 10x knowledge worker
4. Becoming an AI expert

🔮 Understanding AI and where we are heading

AI is not just a technical phenomenon. It is reshaping economics, labour markets, geopolitics, education and the society overall.  Understanding its history and current status helps project (to some extent) its future trajectory. This in turn helps make smarter investment choices, be it actual financial investments or personal time investments, e.g. career choices.

This section is covered by reading a select list of a handful of books that we have found interesting and useful.

Exponential (Azheem Azar)

A book that introduces the concept of the Exponential Gap between technological innovation and societal adaptation. Azheem analyses the evolution of technological innovation in key areas, such as computing and artificial intelligence, renewable electricity and energy storage, and biotech and manufacturing and. Azheem uses the assumption that we are entering a world of abundance across all major resources in order to analyse the world and the future.

✍️ Our comment

This book was a wake-up call for us and triggered an obsession for continuous learning about innovation to stay in touch with a fast-pace changing world.

Superintelligence (Nick Bostrom)

A book about the concept of superintelligence, where machines, due to learning and evolving faster than humans, exceed human-level intelligence. To illustrate this, the author focuses on the emergence of a singleton, a single all-powerful AI with no reasonable competitors. Then, Bostrom analyses the potential threats and solutions to this.

✍️ Our comment

This book helped us define superintelligence in a simple way and appreciate how close and, at the same time, far we are from it.

Life 3.0 (Max Tegmark)

While Superintelligence focuses on AI’s existential risk, this book is centered around today’s big AI debates and near-term risks, such as the impact on the labour market, geopolitics and weapons of mass destruction. In addition, Tegmark moves to more philosophical questions about whether AI can have consciousness and the right to free will.

✍️ Our comment

This book helped us take our AI thinking a step further. Could it be possible that humans may actually not be the most superior beings ever on this planet?

💪 Becoming a conversational AI power user

Large language models (LLMs), such as ChatGPT, are becoming one the main ways to interact with computers for knowledge, research, writing, decision-making and automation. Being any knowledge worker and not knowing how to work effectively across multiple use cases with LLMs in 2025 is the equivalent of being an accountant in 2010 without knowing how to use Excel: you risk being left behind.

Below are a couple of short courses that can help you increase speed and creativity, and become an AI-enabled worker.

Generative AI for Everyone (Andrew Ng, Deeplearning AI)

A free 3-hour online course consisting of many bite-sized easy-to-consume videos that help learn what GenAI can and can’t do, as well as practical exercises across different use cases and modalities (text, image, audio, video).

Generative AI Prompt Engineering Basics (IBM)

An 8-hour online course consisting of videos, practical exercises and a final hands-on exercise that help learn the key elements and different methodologies of prompt engineering, including one-shot, interview patterns, chain of thought, etc. The course is free but it is delivered through the Coursera platform, which does require a 50-dollar-per-month subscription. In any case, Coursera is a great resource of AI knowledge and training courses, and we will come back to its utility later on.

✍️ Our comment

This course helped us realise how important prompt engineering is and the difference it can make in getting what you want out of LLMs. Also, it helped understand the fundamentals of reasoning models (as they use the concept of chains of prompts).

⚙️ Become a 10x Knowledge Worker

AI used effectively and to its full potential can make a knowledge worker 10x more productive, carrying out tasks that previously would take a day within an hour and achieving superior outcomes..

The key phrase is “full potential”. To do so, one needs to use a combination of i) no code AI applications, such as ChatGPT, ii) low-code workflow applications, such as Zapier and n8n, and iii) AI coding applications to develop full code.

Mastering these, a marketing manager can design a Black Friday marketing campaign, draft the content and design the relevant graphics, build the various promotional landing pages, execute the campaign automatically in multiple channels and social platforms, analyse the performance metrics, and receive ongoing alerts on deviations from the original plan. All this, with a few hours of work in a few days and on their own, compared to the status quo approach that is more manual, requires the collaboration with content writers, graphics designers and data analysts and takes a couple of weeks. Can you imagine what a financial analyst or a recruiter or a venture capital analyst can achieve? This is the definition of a 10x knowledge worker.

So, how do you get there?

We differentiate between two AI Knowledge Worker levels: 1) Beginner, and 2) Advanced.

Beginner level

A beginner-level AI Knowledge Worker gets some productivity gains (up to 2x) by automating repetitive tasks, e.g. CRM updates, company research, email summaries, and integrating AI in existing tools.

At this level, AI workers use no-code and low-code tools having some basic, but limited. understanding of what they do, which keeps things simple. The downside is that you won’t really know what’s happening underneath the hood, which means that if something goes wrong, you can’t fix it easily. As a result, it’s probably best to stay away from building anything with a high degree of complexity. Also, at this level AI workers operate at the front end only. This also keeps things simple, as you don’t need to understand how software is built, the differences between front and back end, etc. the drawback is that memory, complex functionalities and security become challenging, if not out of scope altogether.

In any case, it's a great starting point and we recommend it particularly for people without Tech, Engineering or Natural Sciences academic background.

Achieving Beginner-level productivity gains requires familiarity with a combination of:

• A low-code workflow automation tool - we recommend Zapier
• A user interface where basic automation scripts can be added - we recommend Google Sheets, Google Docs and Gmail
• An LLM to generate simple code to embed in the workflow tool and the user interface - we recommend Anthropic Claude
• An information management tool with extensive built-in AI functionality - we recommend Notion

Relevant resources include:

• Zapier: Intro, Use cases
• Google Appscript: Intro, Use case 1: Using LLMs with Google Sheet, Use case 2: Collect Web data within Google Sheet
• Notion: AI tutorial

Advanced level

An advanced-level AI Knowledge Worker can maximise productivity gains (up to 10x) by building recurring end-to-end AI workflows that automate entire deliverables, such as the marketing campaign example mentioned earlier.

At this level, AI workers use a combination of low-code tools and full code created with the help of AI coding applications in order to combine efficiency with flexibility and customisation. While an advanced-level AI Knowledge Worker doesn’t (need to) write code from scratch, they understand exactly how software is designed and built, how front end and back end works, and what the code actually does and doesn’t.

Achieving Beginner-level productivity gains requires a combination of:

1. Understanding software development, in particular 1) coding logic and 2) the basic software components
2. Familiarity with an AI coding application - we recommend Cursor
3. Familiarity with an advanced workflow application - we recommend n8n
4. Familiarity with making API calls to an LLM - we recommend OpenAI’s Responses API

🤔 Coding? Really?

We get asked often about one of the hottest debates in the software engineering world nowadays: “In a world where AI writes all code, does it make sense for anyone to learn how to program software?”. Our view is that “Yes, it still makes sense”.

First of all, there are still some years left before AI can write a piece of complex software end to end that an enterprise can use safely at scale.

But even when we get there, think about mathematical calculations. In a world of calculators and Microsoft Excel, is it valuable to know how to make a multiplication or a division? We think it is, and the reason is the value of intuition. By learning how things work, you develop an intuition that is indispensable in ensuring high quality of outputs. Intuition is what helps a financial manager look at an Excel output and catch an error. Intuition also helps the same manager understand whether the error is in the inputs or in a calculation, and thus identify the root cause efficiently. The same goes with other technological innovations, such as the GPS-based navigation.

Nevertheless, it is indeed still debatable whether you need to study a 3-year Computer Science course or attend a 12-week coding bootcamp and learn specific programming languages in order to build such intuition. Therefore, we provide multiple paths:

1.Learn coding logic in an abstract way

• This can be done by attending an online pseudo-language course. This is effectively like learning how grammar works without learning a foreign language per se.
• Resource

2.Learn coding logic in a practical way

• This can be done by attending an online   course of any fundamental programming language. We recommend Python because it’s easy to learn and has strong synergies with AI/machine learning (linked to following section)
• There are many online courses - see next section for an example we picked.

3.Learn the basic components of software

• This is part of any coding bootcamp or Computer Science degree. There are also specific online courses that cover this topic.
• Resource

The remaining resources that can help reach an Advanced AI Knowledge Worker level are:

• Cursor tutorial
• n8n tutorial
• OpenAI: Official AI agent guide, Agentic API tutorials

🔬 Understanding Deeply How AI Models Work

If you are a geek and believe (like we do), that, in order to fully understand AI and put it into practice, you need to dive deeper than the level you actually need for your daily work, then we got you covered. The below resources will help you understand the limits, biases and costs of AI, as well as give you credibility when working with technical teams.

Deep Dive into LLMs like ChatGPT, by A. Karpathy

A great (and free) starting point is Andrej Karpathy’s (Stanford Data Science PhD, OpenAI founding member, Director of AI at Tesla) 3-hour video “Deep Dive into LLMs like ChatGPT”. There, Andrej explains in non-technical terms how large language models are trained, some core attributes (hallucinations, tool use,  knowledge/ working memory, etc) and post-training actions (finetuning, reinforcement learning, etc.).

🛣️ The path forward

If you want to get your hands dirty, our recommended path is Python → Machine Learning → Deep Learning (which includes Transformer models, which is the technology behind LLMs).

There are many Python courses online. Given that by that point you already have a Coursera subscription, we found an IBM-sponsored course that seems to have a comprehensive curriculum - Python for Data Science, AI & Development. You don’t really need to become a coding expert, but models get trained in computer programs that are written typically in Python, so it’s sort of a pre-requisite. Of course you can try to “vibe code” your way to building an AI model, but it’s worth knowing in depth what you are doing.

We mentioned earlier DeepLearning.AI, an online educational institution founded by Andrew Ng, whose courses are also delivered through Coursera. If you're looking for cost-effective, flexible and well-supported foundational learning, this is a great option.

They have a Machine Learning Specialization, which is a set of 3 online courses that cover everything in ML, from Regression, Classification and Decision Trees to Supervised/Unsupervised/Reinforcement Learning and Neural Networks, and most importantly have hands-on AI model building exercises that get graded with feedback. It takes about 2.5 months assuming you spend 5 hours per week, and comes for free with a Coursera subscription (50 dollars per month).

The next step would be their Deep Learning Specialization, which is a set of 5 online courses that cover everything in DL, including Convolutional and Recurrent Neural Networks, Transformers (which is the architecture of OpenAI’s GPT-3, the AI model that changed the route of AI history), as well as various applications, from text analysis to facial recognition.  It takes about 4 months assuming you spend 5 hours per week, and comes for free with a Coursera subscription (50 dollars per month).

If you are looking for a more immersive, guided experience with strong institutional backing and deeper practical exposure, and you're willing to commit time and money, then Emeritus, an online learning platform, hosts two online courses for postgraduate diplomas from renowned universities: 

1. Applied Machine Learning from Columbia Engineering
2. Deep Learning for AI from Carnegie Mellon University School of Computer Science

Each one takes 6 months to complete assuming 10 hours/week, offer Office Hours with learning facilitators, graded programming assignments, dedicated programme support teams and a final capstone project. If you opt for a course delivered through Emeritus, you get free access to a Python course (which is a prerequisite for both), so you don’t need to sign up to a different one.

✍️ Our comment

We have completed Columbia Engineering’s Applied Machine Learning course and found it very useful, first, because of the helpful faculty that is there to answer questions and, second, because the deadline-driven cohort-based approach helped us maintain a studying rhythm and discipline. Is it worth 2,000 dollars compared to the 300 dollars that a Coursera Deeplearning.AI course would cost (assuming you did it in 5 months)? We are not sure, but we certainly don’t regret doing it. Perhaps the best way would be to try to get it partially or fully subsidised by your employer.

📑 Wrapping up

Our journey through personal development in the AI era has outlined a comprehensive learning path, starting with a foundational understanding of AI's societal impact and progressing to becoming a highly effective AI power user. We then explored how to leverage AI to become a "10x knowledge worker" by mastering various tools and coding principles, finishing with a deeper dive into the underlying mechanics of AI models for those seeking advanced technical expertise.

We think that embracing continuous learning and adapting to the evolving landscape of AI is not an option but a necessity for personal and professional growth (if not survival!). The online training world is huge and the above are just some examples based on our own learning experience. If you have a resource suggestion to make, we would love to hear it.

Finally, given the speed things are moving, we expect this list to become obsolete in 6-12 months, so tune in as we provide updates in the future.

See you next week for more AI insights.