Ship Happens, Week 5: DeepSeek, simple analysis, Perplexity, and building the life you want

DeepSeek, simple analysis, Perplexity, and building the life you want.

Happy Friday everyone! Welcome back to Ship Happens, your weekly product manager newsletter.

I'm using this newsletter to share at least three things l've come across this week to help you build better product. Subscribe so you don't miss these when they come out:

On to this week's thoughts and updates!

1. DeepSeek isn’t the end of “Big AI”

If you’re not subscribed to Understanding AI, I would recommend it because articles like this gem hit my inbox and provide such a better understanding of things like DeepSeek than I will ever get from mainstream news sources or social media:

I don’t believe DeepSeek crashed Nvidia’s stock

AI is evolving so fast, if you work in tech you need to stay up to date on the latest. This newsletter is one of the best ways I’ve found to do that, along with Perplexity which we’ll talk about in a minute.

The author of the article,

Timothy B. Lee

, basically outlines how DeepSeek, while it is a significant leap forward in AI technique and smart leveraging of lower compute power to build a better product than what “Big AI” has built collectively, it’s not the reason why Nvidia’s stock went down this week:

Conventional wisdom holds that Nvidia’s stock crashed because DeepSeek proved that AI progress won’t require larger models and more computing power. I don’t buy this at all.

Prior to this quote he highlights how DeepSeek’s triumph with V3 is more about innovation with a technique called “mixture of experts” which is one of the smartest, coolest innovations I’ve heard about in this space:

• A technique called “mixture of experts” divides a neural network up into a bunch of smaller networks (called experts) that specialize in different reasoning tasks. At inference time, the model automatically figures out which experts are best suited to predict any given token. Only a small fraction of the network’s 671 billion parameters are used for any given token. This reduces the amount of computation required to train or use the model. DeepSeek didn’t invent this architecture, but it made several tweaks to improve its performance.

But ALSO an innovation that is likely replicable by Meta, OpenAI, Anthropic, etc. In fact, friend of the newsletter

Andrew Miller

and I were discussing this over Notes, here’s his provocative note:

And here’s my response that I very much still believe, it’s really about how new techniques are going to pour gas on the proverbial AI fire:

So, DeepSeek is certainly a triumph in AI that is going to push the whole field forward. Does it somehow magically mean that Nvidia chips are worthless? No. In fact, it seems likely that the drop of semiconductor companies are due to new tarrifs, which makes way more sense to me given that these new techniques will allow AI companies to do 10x more with the same powerful compute (helloooo AI agents coming soon?!):

A more plausible explanation is that someone tipped traders off to Donald Trump’s plans to slap tariffs on chips made in Taiwan—which Trump announced later in the day. I can’t prove this theory, but I think it fits the facts better than the DeepSeek theory. Interestingly, we didn’t see a second selloff in Nvidia or TSMC shares after Trump’s announcement, suggesting that markets had already “priced in” the news.

The lessons:

1. Don’t believe everything you see on the news.

2. DeepSeek is good for Big AI, it will push the industry forward

3. DeepSeek is an indicator of even more big things to come with AI + Nvidia chips as the techniques for building models get better and better

4. Follow

   Understanding AI
   and
   Timothy B. Lee
   !!!!

2. Don’t discount what you can learn from simple analysis

The first step in putting my analysis together: staring at an unlit lightbulb…Photo by Diego PH on Unsplash

I have been elbows deep this week in analyzing how users traverse our web experiences on different devices. I’ve been reminded this week that often it doesn’t require a super deep analysis to get great insights that you can use as a product manager.

I’ve gotten so much out of simply comparing behavior on different device types for the same web experience. Just doing a simple comparison of key metrics representing behaviors we want to help users perform showed some pretty glaring gaps that we can start addressing.

So the next time you start thinking to yourself “I need to commission a large, deep study of user behavior across my site”, I’d encourage you to start by opening something up like Amplitude or Mixpanel instead and starting to look at your key metrics in comparison to different form factors and use cases. See where the clear gaps are. And if you don’t have those tools, dust off your SQL skills or grab your data scientist/analyst and start by asking simpler questions first. Save the complicated reports for the deeper questions when you have them.

You’ll go farther, faster with this approach.

3. Perplexity continues to be my AI daily driver

I can’t say enough about how much I love Perplexity. I’ve tried Gemini, SearchGPT, Anthropic, Google Search with AI summaries, and I keep coming back to Perplexity. It’s really amazing at what it does as this article wonderfully summarizes:

How to use Perplexity in your daily workflow 👀

If you haven’t given it a go, it does an amazing job of researching for you with the latest data about any topic you’re interested in. It summarizes, cites it’s work so you can use the links later, and let’s you converse easily to keep the research going.

I can’t go back to Google at this point. I would love more in terms of structured content and discovery, but no one else can beat it!

4. Build the Life You Want

Highly recommend this book that I’m almost at the end of. An excellent guide to building a life of happiness and contentment, something we can all use and appreciate.

That's it for this week!

If you've found this helpful, please consider showing your support by subscribing:

And please consider becoming a paid subscriber at $5/month or $50/year. I'm going to start sharing things like my resume template, resume reviewer for paid subscribers only. So consider going paid to get some of those great benefits.

Finally, share with someone else that you think would find this useful:

I’ll be back around this time next week with more useful product manager things!