Building @exceptionalvc. Ex: @SoftBank | Ex: @Okta | Supporter of @UMichFootball @UMICH
San Francisco, CA
Joined March 2010
- Tweets12.6K
- Following650
- Followers6.4K
- Likes29K
Random rant. Walk with me for a second.
I think delivering bad news on a Friday is a selfish move in business.
Whether you're passing on a founder, rejecting a job candidate, or telling a GP you're not investing after months of conversations, you have four other weekdays to have that conversation.
I get it. Business is business. Nobody owes anybody a yes. But being thoughtful doesn't cost you anything.
What's just another email off your plate could ruin somebody else's entire weekend.
And if you knew you were going to pass on Wednesday, why wait until Friday?
You can be direct without being inconsiderate. How you treat people when the answer is no says a lot about you.
I could not agree more with this take!
This is what VC eating PE looks like.
One physical world and services brand at a time.
VC already had access to tech and capital, but now it has learned to operate and sell in the legacy world, too.
PE has little to no tech depth, but it has access to capital, infrastructure, and knowledge and expertise unique to physical services industries.
Venture previously never touched services. Services was a dirty word. Now it's not.
VC is coming hard for PE. That's the only way to grow AUM massively.
For context, VC is $1 trillion in AUM, while PE is $7 trillion.
Marell Evans retweeted
"run evals to see if performance holds up." anthropic recently released a guide on folding subagents into a single agent loop with skills and tools, and that one line carries more weight than it looks: claude.com/resources/article…
their own example shows why. on a sample support agent, folding one subagent into another cut latency and also broke a process that had been passing: the main agent answered from the order lookup tool without asking the shipping specialist for the delivery date. it only showed up in one run out of two. an average score would have stayed flat.
so their recipe is really a spec for the eval. it needs a task for every chain of agents that actually runs together in production and every rule the agent must not break. each task has to run more than once. and you read it by which tasks flipped, not by the overall number. you can't write that set at a desk. it comes from your own production sessions.
the bigger shift is in the guide's reasoning: each new model lets you move more out of the harness. subagents become skills this quarter, something else gets deleted next quarter. the harness is now the thing that keeps changing. the eval has to be the thing that doesn't. keep a fixed set of real sessions broken out by failure category, keep adding new failures as they show up, and rerun it across prompts, skills, tools, and routing every time you delete something.
your harness will be rewritten. your evals shouldn't be.
Marell Evans retweeted
Congratulations @KeycardAI , @ianlivingstone ... GO TIME!
The 4th annual Cyber 60 list is here.
This year’s list recognizes some of the fastest-growing venture-backed cybersecurity startups in the market, building to advance next-gen infrastructure, AI security, and how we approach threat detection.
Congratulations to everyone who made the #Cyber60 list in partnership with @FortuneMagazine and sponsor @AWSstartups.
Read our full CISO survey report, conducted with Wakefield Research, which polled 200 CISOs on enterprise AI and where they’ll invest next: lsvp.com/cyber60-2026-2027/
Marell Evans retweeted
What a team! 🫡🚀 @trajectorylabs
Marell Evans retweeted
Let's goooooo, so excited to have you on the team 😀
I've always thought putting Continual Learning in people's hands is as much a design problem as it is a research / engineering problem, and we need the best of the best to will that into existence
some personal news: I joined @trajectorylabs to build the interface for continual learning :)
we’re building a future where anyone can train a model and have it learn from the real world
so excited to be a part of this team and bring this vision to life!
Marell Evans retweeted
We're announcing the public beta of datavendor.ai
We spent the last two years building the best solution for people buying and selling data, whether you're a researcher in a lab or a company trying to sell data.
Data is the fundamental bottleneck for AI progress, and DataVendor will be the infra that solves it.
I can put together a plan.
Send it to me.
Contract buyout for Michigan football OC Jason Beck is hefty freep.com/story/sports/colle…
Marell Evans retweeted
reliability is what matters at scale. As we already reach abundance of intelligence for most economical tasks - reliability is what will matter even more!
Evals is the biggest bottleneck for enterprise adoption
Databricks Chief AI Scientist Jonathan Frankle (@jefrankle) says the bottleneck to enterprise AI adoption is evals, not models.
"The models are good enough at a lot of really complex enterprise tasks and knowledge tasks"
"the real challenge is describing what you want AI to do and knowing whether AI did it"
Marell Evans retweeted
I’m hiring investors to lead our gtm, partnerships and strategy!
Feel free to dm me
Marell Evans retweeted
Michigan football fans deserve better. Better coaching, better kicking, better play calling.
Marell Evans retweeted
FIRE JASON BECK
FIRE JASON BECK
FIRE JASON BECK
FIRE JASON BECK
FIRE JASON BECK
FIRE JASON BECK
#GoBlue
Marell Evans retweeted
Most people who dream of a quiet life without work are just exhausted from work they hate. The reason you get bored on vacation after a week is that you are meant to move, pursue something exciting, and engage your body and mind in something useful. Do whatever it takes to find work that energizes you. It may take years, but it will have the most meaningful effect on your quality of life.
Someone put me in touch with Jason Beck …. I am going cover his flight back home.
Let’s get him the fuck out of Ann Arbor.
Man, I’m so tired of the negative Nancy bullshit every time some new technology drops.
Every damn time, it’s the same conversation. Jobs are gone. The economy is cooked. The world is ending.
Like damn, can we talk about what’s possible for once?
Nobody’s saying there won’t be challenges. But imagine how many businesses we can build, how many people we can empower, and how many new opportunities we can create.
Our job is to push humanity forward, not sit around scared of the future.
And I’m sorry, being pessimistic about everything doesn’t make you intellectually superior - you look more like a loser.