Hi I'm Morgan, and since I have a lot of new followers, I thought I'd take a moment to introduce myself 👋 First, please don't mistake me for an expert, because I am far from one. I am, however, someone that gets excited about a lot of things, and I'm endlessly curious. I was born and raised in Berkeley, CA, went to college and studied Engineering in Pittsburgh at Carnegie Mellon, and after school went to work for a tiny startup called Sonos. When I joined Sonos, they were pretty much running out of money, I was actually told I might not have a job for long. But I loved the idea, really liked the founders, and took the plunge. Luckily, Sonos did not run out of money, and I was there from ~35 people to ~1,000 people. I was at Sonos for nine years, but knew working for a company, esp. one that was now 1,000 people, wasn't my path, I really wanted to start and run my own business. After that, I started Bold Metrics which I run today as the Cofounder/CTO. I have an amazing team that I am incredible proud of. They are all 10x smarter than I will ever be. Through this journey I realized that my passion in life is coming up with ideas, building the narrative around those ideas and how they fit into a future that I can see in my mind, and pitching those ideas to investors. From there, I can do what I love the most, building teams. I really enjoy the process of finding great people and building a culture where they can truly thrive. At Bold Metrics, I pushed us to adopt agentic coding workflows in early 2025. Through this experience I started benchmarking our own stack, actual PRs from our repos, and helped my team understand what models and effort levels they really needed for regular daily work. What I learned is that benchmarking models across effort levels led to fascinating results. As I shared this with a few friends in SF, they all said I should open source my benchmarking tool and share the results. So I did, and I called it VulcanBench. Today VulcanBench is the biggest money-losing thing I've ever done, and I love it. I get to wake up every morning and see new insights about models that I try to turn into real signal that other engineering teams can use. I live in Lake Tahoe, it's my favorite place on earth. I'm not a big foodie or restaurant guy, and I don't drink, so the whole "dinner and drinks" thing isn't really my jam. I would much rather chat while riding a ski lift then in a noisy crowded restaurant. And I love cooking so would always rather cook at home. For decades now I've been traveling Internationally. I got the travel bug launching Sonos in Asia, Australia, Latin America, and Canada. I usually visit 4-5 countries a year. It's hard for me not to get antsy being in the US for more than a few months, the world is my home. I married my best friend, the love of my life, we've been together for over 20 years now. To me, the meaning of life is love. I live my life by one simple rule - spend as much time as possible with the people I love. You will never find me missing dinners to sit in an office grinding away while eating DoorDash. Time is too precious for that. I am dedicated to changing the world, that is my life mission, and I have known it for a long time. Throughout my life I've had many friends say I'm crazy for saying that. That's okay, we need more crazy people in this world with big dreams and a high tolerance for risk. If you are following me, and actually like and comment as yourself, thank you. It has been a truly humbling experience to see my account on X grow as I just write what's in my head on here. I do not write sponsored posts or do any kind of sponsored stuff at this time so please don't pitch stuff to me, you'll always get a no. I want to keep this account focused on me just writing whatever I want, whenever I want. Okay, this turned out longer than I expected, so I'll leave it here. And holy moly, if you read all of this, you might just be as crazy as me 🤪

Soooo, for anyone on MacOS that wants to go a step further in securing things, there is one patch beyond your browser and other apps that can't hurt to try. If you run this: /usr/bin/curl -sSo /dev/null -w 'built-in curl: HTTP %{http_code}\n' dv-g3b-emudhra.we-have-a-pro… and get a 200 response, that means thing like git, Python, etc. are verifying certs without checking revocation. You can fix this by just pasting this in the terminal: sudo cp -p /etc/ssl/cert.pem /etc/ssl/cert.pem.bak && sudo awk '/^(###|===) /{skip=0} /^=== .*\/CN=(AAA Certificate Services|emSign ECC Root CA - G3)$/{skip=1} !skip' /etc/ssl/cert.pem.bak | sudo tee /etc/ssl/cert.pem > /dev/null Probably overkill for most but wanted to share either way!

Great summary of today, and better to try saying it, out loud, in one breath if you can.

Claude’s personal agent is here, and I gotta say, I love the name, Duet.

One of the best interviews I’ve seen w/Sam.

I have been donating to BCR for almost a decade now, such a great organization. Great raffle to participate in if you can!

I was in Europe for ten days. Tried to compare Dot, Muse, and Grok Bot along the journey. Dot was unfortunately unusable on the trip, both Muse and Grok Bot worked the whole time. But Grok Bot was much more useful and proactive. Muse often responded to messages with things like, “Glad to hear it” and “Logged” Grok Bot, on the other hand, would provide much more detailed analysis and recommendations, it really was like having a travel concierge with me for the whole trip. Just broke the news to my Dot.

So this may sound ridiculous, but lately I’ve been finding building eval suites, and the window they give me into the current frontier as a form of art. This has been on my mind quite a bit lately, about how for me, my creative act in life right now isn’t painting or making music, it’s building evals and waking up every morning to new data about how a model thinks. Seeing the traces, and understanding how a new frontier model approaches tasks that I give it, has become an art form to me. And in some strange way, this all really came together for me last week, when I visited Dali’s home on the coast of Spain in Port Lligat. Yeah, I know I’ll lose some people here, trying to compare evals suites and Dali, but whatever, I’ll just share because that’s what I do on here. When I was walking around Dali’s house, and learning more about his life I learned something interesting that I didn’t know. I always thought of Dali as this incredibly eccentric artist and socialist, someone who was the life of the party. But as I walked alongside his pool, with a couch shaped like lips with Pirelli tires logos around it, I read more about Dali and his parties. At this wild and unique house, along the rugged seaside of Costa Brava, Dali did indeed throw amazing parties, but he himself wasn’t necessarily the life of the party. Instead Dali enjoyed being an observer, watching his guests interact and enjoy the weird and creative world he had created. For him, in many ways, these parties, and more specifically, watching people interact and studying how they thought, was something he saw as fascinating, and in many ways, a form of art. And yes, as I was walking through Dali’s pool area, by the lips couch, with statues of the Michelin man across from me, I was thinking about LLMs, and how models think, and how to me, at this point in my life, this is my art. But not only that, in some ways, like Dali observing his guests at his parties, interacting in the environment he created for them, I saw a strange parallel, with me observing frontier models thinking, in the environments I create for them. Art is a beautiful thing, and in many ways, what makes art so special or unique is the range of medium that it can embody. We often think of Dali for his amazing paintings, and of course these are some of the most spectacular forms of art ever made. But I don’t think many appreciate the art Dali made, at his house on the coast of Spain, an art not made by paint brushes, but by people, interacting and the beauty of how humans think. As new frontier models come out, and I continue to build more interesting and complex evals suited to challenge them, I can’t help but find a parallel here. And while I’m no Dali, maybe in some small way, I am entering a more creative, artistic point in my life, just in my own, unique and different way. The first photo below I took last week, the second is a piece of history. Morning brain dump complete. Thoughts?

Super interesting workflow, prompting a strong planning model like Astra to run code via a CLI like codex on worktrees vs relying on subagents.

Must read on using decision models for evals from Hamel

Another System One model has entered the arena 👀 Everyone keeps talking about decision models and comparing them to Jev. But most decision models are still LLMs at their core, not System One models like Jev. This could mark a turning point. And from Cloudflare who is absolutely on fire rn, wild.

Super cool stuff from Artificial Analysis 🔥

New benchmark coming, keeping this one a secret 🤫

this is insanely cool

The traces for my Muse Wonder Full benchmark with @VulcanBench are now live on Github. I'm a big believer in not just sharing data and results, but also showing all the details so anyone can analyze every single result. To me, the most interesting thing about looking at traces is seeing how different models think I find Muse Wonder Full to actually be a very interesting model to see think because it's actually very accurate, but overthinks, causing it to use a lot more tokens, and take a lot more time than other frontier models. To me, I don't really see this as bad, more like an opportunity to see a model before it reaches its full potential in many ways. Lately I've been thinking of frontier models, and more specifically, the traces from these models, as a form of art, capturing a moment in time, a month or two later, these models will be completely different, but the traces will always be a piece of history. Maybe some day we'll have the VulcanBench museum where you can look at early model traces and examine them like ancient art...haha, okay, maybe only I would like that. Enough rambling, here's the traces for anyone who wants to do a real deep dive. Live long and benchmark 🖖 github.com/morganlinton/vulc…

Good analysis and similar to what I saw with VulcanBench. Overall, Muse is very promising, just seems like there’s an opportunity to fix token use/overthinking issues.

Super cool idea, and great name