Okay well it looks like I picked the right week to start my newsletter about AI + coding because so much has already happened that I realized I just need to ship this newsletter before any more news hits!

First, welcome to my newsletter, the goal here is simple - I’m the cofounder and CTO of a startup (shameless plug for Bold Metrics), and I’m fanatical about how AI coding tools are changing the way we build software. I can see the impact these tools are having on our productivity, and on the world, and I can also see the pace that the space is moving, and that pace can only be described as meteoric.

I’m drinking from the firehose when it comes to learning about and testing out new AI coding tools and I’ve been taking notes like crazy. Lately, whenever I talk to friends I can’t help but go on and on about some new AI coding tool I’m trying out or some wild feature someone added that makes an existing tool even better.

And I guess I talk about this topic so much that I’ve now had more than a few people say, “why don’t you write a newsletter about this??” So I decided to give it a shot and here we are.

What I hope will make this newsletter different from any other newsletter out there is that I’m someone who isn’t just reading about these tools, I’m actively using them, putting them through as many tests as I can throw at them, and having my team do the same. Through this process I’m figuring out what’s real, what’s hype, what’s working, what’s broken, and continuing to try out new tools the moment they come out.

Some people have a “way they’ve always done things,” and I’ll just cut to the chase, I’m not that guy. I believe we are moving into a time where engineering teams that leverage AI coding tools will be able to innovate faster - it’s that simple. And those teams that stay on top of the ever-changing AI coding landscape have the best chance of staying on the cutting edge. So welcome, thanks for reading my little intro here, and now - let’s get to the good stuff.

Bolt announces $105.5M Series B led by Emergence & GV

It’s no secret that Bolt has been one of the fastest growing companies in the AI coding space, and today the company has a lot more wind it its sails as it announces a massive $105.5M Series B round led by Emergence and GV.

If Bolt isn’t on your radar, hopefully it is now, and I’ll give you a little overview of it for anyone who hasn’t been Bolting away like I have for the last few months.

Intro to Bolt

I always explain Bolt to people like this - tools like Cursor and Windsurf are for software engineers to leverage AI in their existing development process. Bolt is like having a access to a full stack engineer that can build complete applications. I’ve been using Bolt for so many things from prototyping to end-to-end development, and as a way to fast-track my own knowledge of libraries like Tailwind CSS.

A few weeks ago Bolt added a really solid integration with Supabase which made database integration completely seamless, something that many AI coding tools have struggled with.

And Bolt keeps adding new features, almost by the day at this point with a really cool new announcement this week that you can see in the tweet below:

If you don’t know about lazy code omission, it’s a real problem that LLMs face when writing code and something that many of us have had to just learn how to adapt to. The concept is simple, rather than actually writing the code. Here’s the dets in Bolt’s own words:

This came on the heels of another update Bolt did, also solving for a common issue that LLMs face around hallucinations when rendering icon names:

I could go on and on about all the innovation Bolt is bringing to the table but I’ll stop here for now. Suffice it to say, even before announcing the $105.5M fundraise Bolt was innovating at an amazing clip, now I can’t even begin to imagine what’s ahead for 2025.

DeepSeek R1 takes on o1

Two days ago, Chinese startup DeepSeek announced the release of their newest Open Source model - R1, and dropped one sentence that nobody was expecting:

“Performance on par with OpenAI-o1”

Before I dive in, here’s the tweet that started all the hullabaloo:

First things first, I think we can all agree, OpenAI o1 is an incredibly powerful model, and after releasing their $200/mo Pro Plan people have been using it to do some pretty interesting stuff. In fact, people are using the Pro Plan so much that Sam Altman announced that the company is actually losing money on the plan, an unexpected but not altogether bad outcome. 

Now, an Open Source model, backed by an MIT License that allows people to use it commercially with pretty much no strings attached, is putting the pressure on o1, a model that really was in a league of its own just last week.

Okay, so let’s dive into the good stuff, here’s a deeper dive into some benchmarking:

As you can see from the last line, DeepSeek-R1-Distill-Llama-70B beats o1-mini in every test with the exception of CodeForces where they’re within 10% of each other.

And if you really want to geek out, DeepSeek put out a pretty detailed technical brief on the new model, here’s the TLDR; 

Oh and when I say that the technical brief is “pretty detailed” I mean like PhD-level math detailed because this is actually in the brief:

If you can understand that, you should stop reading this right now and go build a foundational model, seriously, don’t waste your time here 🤪

For me as an engineer I’m incredibly fascinated by the fact that we now have an Open Source model that’s on par with o1, I think this opens the door to a lot of really interesting innovation opportunities. While OpenAI’s Pro plan at $200/mo isn’t that expensive relatively speaking, let’s be honest, for most people adding a $200/mo expense just isn’t something they can do.

The more we lower the barrier to entry for engineers to play around with things, the more likely they are to stumble on something they wouldn’t have thought of before. In many ways I see us moving into a time where the opportunity for serendipity is higher than ever before, and it’s because engineers don’t have to work for a company, that pays for some expensive tool or license, it’s because everyone can use these tools to build.

Okay, but let’s talk about the elephant in the room, this new LLM thrusts us right back into the TikTok debate in many ways and it’s safe to say the US news isn’t going to miss a beat here.

This is going to be a really interesting one to follow and I’m already starting to play around with R1 so will hopefully have some more examples from my own person experience in the next issue!

Windsurf Wave 2 wows the crowd

While Cursor has been the star of the AI coding IDE world, Windsurf has quickly gained steam and has done a stellar job of innovating quickly and adding features that developers are really digging.

And it’s safe to say that with their newest release, aptly called “Wave 2” it could be pulling ahead of Cursor. Here’s the tweet that kicked off all the excitement:

Yes, you’re reading that right, Windsurf added functionality over the weekend that allows for agents to actually use the Internet to research and make code changes based on what they find. Santiago did a great job summarizing the killer features in Wave 2:

The last sentence in Santiago’s tweet is one that I think everyone reading this should think about - let this sink in:

I don’t think I remember what writing software without AI looked like. I don’t even want to think about it.

Now think about where you are today. If you’re still coding the old fashioned way because, “that’s how you’ve always done it,” just think about how much longer you want to hold off, because the world is moving fast and expectations around output and velocity are changing.

The three stories I covered in this issue are three of probably thirty major topics I could have covered in the AI coding space, that’s how fast it’s moving. What’s so interesting about this change is that the end result isn’t AI replacing humans, instead it’s unlocking human potential and allowing engineers to write code and ship faster than ever before.

Bonus Nugget - TikTok’s parent company just released an AI-powered IDE 👀

So this one just landed on my radar, and I haven’t had a chance to play around with it yet…but I’m definitely taking it for a spin this week. Bytedance, TikTok’s parent company released their own AI-powered IDE to compete with Cursor and Windsurf, it’s called Trae and it looks pretty slick.

Expect to see a more detailed review of Trae in the next issue of my newsletter.

Okay, so I told myself I’d try to stick below 1,500 words, and I’m above that now so that’s it for this issue. If you read all the way to the bottom, thank you, if not, then you’re not reading this anyways 🤪

See you next time, Issue #1 is officially fin.