In an AI world, do you really need Excel, or Word or PowerPoint?

How does Liberty Ventures evaluate deal flow? In last week’s Angel Investor Accelerator, our Chief of Staff @soymichelleber shared a key shift in how we think about founders: It’s not just, “Can this founder build a large business?” It’s, “Do we trust this founder to lead one?” That lens shapes how we source opportunities, assess character, and build conviction before investing. For investors who think this way, we’d love to connect. Join our network at libertyventures.vc/

"Always be honest" sounds great in theory. For humans or for AI. But reality is messy. Truth changes. Context matters. And pure honesty without empathy can be cruel. Effective communication requires more nuance than simple "truth." For humans or for AI. longform.asmartbear.com/fail…

🚨 THE ROOM WHERE IT HAPPENS What if one connection could change the future of your business? At the Principled Business NYC Summit, the most ambitious investors, executives, and founders come together to build, fund, and scale the next wave of freedom-driven innovation. 🚀

Last week in our Angel Investor Accelerator, we heard from @seanthinks, CEO of EasyBot Chat. Most AI tools are built for desk workers, but the majority of the workforce is on the front lines, i.e. in clinics, kitchens, stores, warehouses, and beyond. EasyBot Chat is helping bridge that gap by giving frontline teams mobile access to company knowledge in their preferred language. A thoughtful conversation on where AI can create real operational impact. Learn more: libertyventures.vc

An easy way to get unstuck is to get up and take a walk. We generate more creative ideas during and after walking outdoors—and even on a treadmill facing a blank wall. Divergent thinking rarely happens when we're tethered to a desk. Moving our bodies frees our minds.

Most A/B tests are impactful enough to matter. Even when they’re “positive,” at least 50% of the time they are 𝘯𝘰𝘵 positive. “Why not!?” you say. “Isn’t it math?” Yes, 𝘵𝘩𝘢𝘵 is the math. Come see: longform.asmartbear.com/p-ha…

every founder is racing to ship faster. nobody is racing to operate better. the moat moved while we were all watching build speed.

i run a $2.5M arr ai agent for upwork. here's what shocks me about other products in this space. they're obsessed with ai as a feature. they think the next model upgrade will move the needle for their customers in lead gen. they end up with unpredictable outputs and 25% monthly churn. meanwhile our fine-tuned, eval-ed-to-death, guardrailed gpt-4o beats their opus 4.7 on the same tasks. an ai agent is only as smart as the team building it. what we actually do to bulletproof our ai: - hand-labeled benchmarks for every small ai completion - multi-turn evals for every agent tool - 500 regression evals per agent, before any prompt change ships - user feedback on every ai turn, wired back to evals - langfuse for production span tracing human labeling - support team trained to add new evals from real tickets - adversarial prompts for every guardrail - prompt template versioning in source control, diff-reviewed - cost latency evals, not just quality - offline eval gate blocks every deploy on regression - canary cohort fast rollback on any eval miss - weekly bake-offs (fine-tuned 4o vs sonnet vs opus) on the live benchmark the "another ai feature" race is a vanity loop. discipline beats novelty. what's your team's regression eval count?

1,000 cold emails vs 1,000 Upwork bids. Same founder. Same month. Here's what actually happened: → 1,517 cold emails: 5 positive replies. → 1,000 Upwork proposals (automated with GigRadar): 111 positive replies → $80k/mo in retainers closed. The wild part? @emaildeepdive runs an email marketing agency. Email is his product. Here's why the gap is that wide, and it has nothing to do with deliverability or copy: Upwork clients are problem-aware. They already know they have a problem, they've decided to spend money on it, and they're actively looking for someone. You're not selling them on the problem, you're selling them on you. Cold email is the opposite. You're interrupting someone who wasn't thinking about the problem five seconds ago, convincing them they have one, then convincing them to solve it, then convincing them to solve it with you. Three sales in one sequence. No wonder reply rates look the way they do. Gavin's take for anyone starting out or running lean: lock in Upwork first. It's cheaper, faster, and the leads are warmer by default. Layer cold email on top once your inbound engine is already humming. Which is basically the thesis we've been betting GigRadar on for three years. Good to see someone outside the bubble run the numbers.

Mark Cuban said what every CEO in America needs to hear. "There are only two types of companies in this world, those who are great at AI, and everybody else. If you don't know AI, you are going to fail. Period, end of story." @mcuban didn't stop there. "Whether you are an employee, you're going to have to understand how AI impacts your job and how you can use it to be better at your job. Same if you're a student. And if you're a CEO, you can't just say, okay, I'm going to get my tech guys to understand it and educate me on it. You have to understand it yourself because it will have significant impact on every single thing that you do. There's no avoiding it." This is coming from someone who built and sold the Dallas Mavericks, who made his first fortune selling a company to Yahoo for $5.7 billion in 1999, and who has spent the last two years going deeper on AI than almost any investor his age. He is describing the Innovator's AI Dilemma, entrepreneurs are right now building AI-native companies designed to displace every major incumbent. If a CEO tears down their company to rebuild it AI native, investors revolt but if they do nothing, AI-native startups eat their market and investors revolt anyway. Either path leads to shareholder lawsuits and there is no comfortable middle. The companies that survive will be the ones where leadership, not just the tech team, genuinely understands what the technology can do. The ones that don't will look back at this moment the same way Blockbuster looks at 2005.

Big news 🎉 Pouncer.ai's Profile Optimizer for Upwork has been acquired by @GigRadar_io. 5,000 freelancers used it to sharpen their profiles. Now it's part of a platform whose users have generated $45M in Upwork contracts.

Once you start to implement Claude Cowork (or OpenClaw) in your daily tasks, you quickly realize the huge role of #MCP in AI. If you have an API for your product, then you must have an MCP server if you want to stay relevant to your customers.

For anyone using Ai tools like Claude or Gemini or ChaGPT for getting legal advice…be careful! Recent ruling has determined they are NOT protected under attorney/ client privilege and responses can be used by the government (and also -assuming-as part of civll discovery process) omm.com/insights/alerts-publ…

I loved recording this podcast with @lennysan about how do diagnose “why aren’t we growing faster,” and of course what to do about it. Come hear it on Lenny’s Podcast, including on YouTube: youtube.com/watch?v=8xLquwfx…

ICYMI: Fresh new article dropped: What you decide to stop doing, or never do, gives you the space to do whatever is better. And there’s 100x more of those (without exaggeration). So, you could start there, and start deleting. longform.asmartbear.com/focu…

Just published a new article! 👇🏾 Do less, to create the space to be great. Many thanks for pressing 🔁 and ❤️! longform.asmartbear.com/focu…

Instead of seeking feedback on how you did, it’s more effective to seek advice on how you can do better.

🚨 The Upwork Research Institute’s new Future Workforce Index Report just dropped 🚨 Check it out to learn how freelancers are shaping the future of work and leading the charge in AI innovation: spr.ly/60112ZmCc

Just learned of the passing of Pope Francis. He was a good man and great Pope whose love and compassion was a source of inspiration for hundreds of millions around the world. May he rest in peace.