What I Learned in 2024

Hey everyone,

Welcome back to Stocks To Space, where I curate the best ideas, tools and resources I’ve found each week as I explore my curiosities.

This is my final post of 2024, so I’ve designed each section to zoom out and focus on the year in review.

Expect favourite picks from my own content this year and pieces of wisdom that will stand you in good stead for 2025. I hope you enjoy it. Speak to you all next year!

As always, if you find something thought-provoking, forward it to a friend.

IDEAS
What I Learned in 2024

This is the second year in a row I’ve conducted Sahil Bloom’s well-known Personal Annual Review exercise.

Truthfully, it has been one of the most valuable things I have done the entire year.

To quote Sahil directly on why it’s so important:

“The end of the calendar year presents us with a unique and valuable opportunity: To reflect on the year that was and plan for the year that will be.

In our rush to look forward, it's easy to glaze over the former and focus on the latter—but a failure to reflect will eventually result in a failure to grow.

It has been a transformative exercise—one that has had an outsized impact on my personal progress and growth.”

After two years of doing this I can already feel the transformative effects of it.

So much so that I wanted to share the 24 (condensed) lessons I learned in 2024, which arose from my Personal Annual Review. 

1. The (hard) things worth doing always suck in the beginning but are so rewarding once done.

2. Your fear, uncertainty, and doubts feel productive but are useless in achieving your goals.

3. Self-doubt and impostor syndrome are a feature of an ambitious mind, not a bug.

4. Your permanent physical location, environment, and context matter way more than you think.

5. Changing your environment (walks, nature, new workspaces) periodically is the easiest way to spark creativity.

6. Shorter, hyper-focused to-do lists lead to better execution versus sprawling lists of goals.

7. Using AI and outsourcing grunt work isn’t “cheating” but strategic leverage.

8. You suffer more in your imagination than in reality.

9. The ability to multi-task is a strength for founders.

10. Strong family relationships provide stability and a long-term perspective for your goals.

11. The narrative in your head can be the biggest thing holding you back—guard against its self-sabotage.

12. Carefully choose your circle—limit contact with people who drain energy or belittle ambition.

13. Nothing focuses the mind better than pursuing one goal relentlessly for a long time.

14. The easiest way to build desired behaviour is to do the thing once a week for a year.

15. Realism complements ambition—acknowledging the risks of failure will bolster, not weaken, your drive.

16. Most meetings and connections waste time—be ruthless in protecting attention.

17. Consistent daily systems, even minimal, compound significantly over time.

18. Removing distractions (alcohol, needless socialising, doom-scrolling) is the only way to free up attention for real work. This is a constant battle.

19. Discarding or adjusting goals mid-year is okay if they no longer serve you.

20. Reading non-fiction raises your floor (knowledge). Reading fiction raises your ceiling (imagination). Both are important.

21. Sometimes, just getting to your desk is the best productivity hack. Don’t make getting to work more complicated than necessary.

22. Controlling what you can is more sustainable than fixating on what you can’t.

23. Life is too short, no matter how old you are.

24. Sometimes, you have to go with your gut. And that’s okay.

If you’d like to conduct your own Personal Annual Review (which I highly recommend), here is a helpful guide on each question and how to answer it.

(Let me know what your favourite lesson is on X/Twitter.)

RESEARCH
How’s Your Model?

2024 was the year for capitalising on AI hype. Big Tech hyperscalers (Google, Amazon, Meta) and startups (OpenAI, Anthropic, Mistral) relentlessly came at us with AI product release after AI product release.

My favourite tool of the year for keeping up with model performance and benchmarking them against each other was Artificial Analysis. Its interactive charts are the best way to establish who is who in the AI zoo.

For me, this method of benchmarking models is much more palatable than the whiplash experienced from following the ever-changing LLM Leaderboard—one week, it’s GPT-4o on top, then it's o1-preview, then it’s Gemini.

Adoption’s Sobering Reality

While AI companies capitalised on the latent hype, the rest of us had to deal with the reality of figuring out how to use these tools in our businesses.

If 2023 was the wild night out celebrating ChatGPT (complete with grand visions, bold predictions, and endless rounds of “AI will change everything”), 2024 was the inevitable morning after.

Corporations traded their euphoric predictions for pragmatic implementation plans and learned that turning those 3 AM epiphanies into real solutions requires more than enthusiasm.

This has been a tough pill to swallow—even though 77% of companies are already using or actively exploring AI technologies across their operations, these solutions are primarily demos. Corporations are struggling to get anything meaningful into production.

However, this is natural for any genuinely disruptive technology like AI.

Jeff Bezos recently compared it to being as revolutionary to the global economy as electricity. This might be the most reasonable comparison regarding the adoption curve.

Adoption will take time because corporations, the primary adopters of electricity and now AI, will proceed with far more caution (but have a more significant economic impact) than the average consumer.

And although adoption is dragging, let’s face the facts:

• Models are getting more intelligent, faster and cheaper.

• Infrastructure is more reliable.

• Tooling is getting better.

And this is all happening at a rapidly increasing rate.

There are opportunities all over the place to create value using AI.

There will be many losers but a few big winners.

Ultimately, the opportunity will go to those willing to suffer through the longer-than-expected adoption curve.

AI Word of the Day

Scaling Laws

In artificial intelligence, scaling laws describe predictable relationships between a model's size, the amount of data it is trained on, and its performance.

These laws are typically expressed as mathematical power-law relationships, showing how increasing key factors—such as model parameters, dataset size, or compute power—can improve capabilities like accuracy, language understanding, and generalization.

INSIGHTS
1 Article

DeepSeek-V3, ultra-large open-source AI, outperforms Llama and Qwen on launch

Speaking of models, China just stole the show with what might be the final model release of 2024.

DeepSeek, a Chinese AI startup, released DeepSeek-V3 this week. The open-source LLM uses a mixture-of-experts architecture and has 671B parameters. Only 37B parameters are activated for each task, allowing the model to process information at 60 tokens per second, 3x faster than previous versions.

The craziest thing about DeepSeek-V3 is that it outperforms Meta's Llama 3.1 and other open-source models, such as the Chinese model Qwen, in benchmark tests.

If that wasn’t good enough, it matches or exceeds GPT-4o on most tests. Surprisingly, only Anthropic's Claude 3.5 Sonnet consistently outperforms it on specific specialised tasks. The company reports spending $5.57 million on training through hardware and algorithmic optimizations, compared to the estimated $500 million spent training Llama-3.1.

Luca’s take: With all of that said, there seems to be a stubborn belief that the US leads the AI model race by far. This couldn’t be further from the truth—model releases like Qwen and DeepSeek-V3 prove that. I believe that China’s progress in AI is grossly underrated and has to be appreciated by the Western world. After all, consumers and corporations vote with their dollars—if China makes the best AI products, the world will choose them over others.

1 Post

• 2024 was the year of multimodal AI models.

• 2025 will be the year of AI agents.

1 Video

TOOLS
My Top Picks For 2024

Unsurprisingly, my top 3 favourite tools of this year are AI tools…

Above everything else I tinkered with, these delivered real value in my daily work.

1. Perplexity.ai — For research and tailored queries that need verified sources.

2. ChatGPT — For resource-intensive queries and workflows, using o1.

3. Claude.ai — For creative writing and coding.

(Let me know what your favourite productivity tool is on X/Twitter.)

RESOURCES
My Top Picks For 2024

Newsletters

• What I Read This Week by Chamath Palihapitiya — For general reading links.

• Bay Area Times — For business, finance and tech news.

• Mark Manson — For personal growth.

• Daily Stoic — For timeless wisdom.

Podcasts

• All-In — The backchannels of Silicon Valley.

• Bg2 Pod — An investor’s perspective on technology.

• Y Combinator — Startup knowledge, insights and best practices.

Courses

• Maven — the one place professionals should go to grow their careers.

(Let me know what your favourite resources are on X/Twitter.)

THOUGHTS
Quote I’m Pondering For 2025

“Pain plus reflection equals progress.”

— Ray Dalio

(Click here to share this on X/Twitter.)

Was this email forwarded to you? If you liked it, sign up here.

If you loved it, forward it along and share the love.

Speak to you in 2025,

— Luca

P.S. Never invest in the present…