Nick Caldwell

Just dropped on The Rebooting with Brian Morrissey. We talked about the transition from media’s industrial logic—scarcity, gatekeepers, mass production—to platform logic: abundance, algorithms, iterative optimization. When everyone can publish, the bottleneck shifts upstream to distribution, incentives, and the orchestration layer. One line that stuck: the creator economy can feel like a content casino—millions of dependent suppliers, a few breakout winners, and constant pressure to optimize for what the machine rewards. AI accelerates that dynamic by resetting the floor. “Good enough” gets cheaper and more abundant, fast—which raises the bar for the middle tier and concentrates leverage even more in distribution and incentives. Link in comments.

4

1 Comment

Sharing a quick update on our work at Zeno AI. We’ve been rethinking how robots move and think. Our latest paradigm-Zero-1, moves away from the "stop-and-act" decoupled control of the past. By using Whole-Body Coordinate (WBC) learning, we’ve achieved a level of synchronization that allows for much more contact-rich tasks. Hanging a shirt might look simple, but for a robot, coordinating two arms and a mobile base while handling soft fabric is a huge challenge. Zero-1 handles the dynamic tension and spatial coupling seamlessly. It’s a big shift from tabletop execution to true whole-body coordination in complex environments. Excited to see where this leads us next. #Zeno #WholeBodyCoordinate #PhysicalAI #MobileManipulation

9

This conversation cuts through the hype to examine why looking at observability initiatives from the frontend isn't just another tech industry fashion statement. https://bit.ly/4dWwrda

4

What Gowri said! OpenAI just shared its latest work on real-time voice. A simple housing conversation rarely stays simple. “Find me homes I can afford.” “Actually, only ones with a shorter commute.” “What about this one?” “Can I see it Saturday?” That sounds natural to a person. For an AI system, it is a lot to keep straight: intent, context, tools, timing, and what changed along the way. That is why Gowri’s point about reliability matters. This is where a lot of conversational AI systems struggle today. The underlying models are improving quickly. The harder part is turning that into something that holds together in a real workflow. There is a lot of work behind getting that to behave consistently. We are still early, but this kind of progress starts to unlock what is possible. It’s been great to be part of a team doing some of the most interesting work in this space in housing right now Zillow   #AI #VoiceAI #AgenticAI #MachineLearning #ConversationalAI

37

3 Comments

Software you ship is mostly software you didn't write. A typical production app pulls in hundreds of third-party packages at build time. Each one is a path into your environment, and most have never been reviewed at the depth your own code has. That's the part of the risk picture shifting fastest right now. Anthropic's Mythos and OpenAI's GPT-5.5-Cyber are the visible examples. Mythos surfaced bugs in OpenBSD, FFmpeg, and the Linux kernel that survived decades of human and fuzzer review. The operational story isn't the headlines. It's that reasoning models catch a different class of bug than fuzzers do. Fuzzers find what mutation can reach. Reasoning models find what requires understanding intent. The bugs in your dependency graph that survived a decade of fuzzing didn't survive because they were hard. They survived because nothing was reading the code with intent. That capability is now available to defenders. It's also available to everyone else. The exposure isn't in the popular packages. Those get scrutiny. It's in the transitive dependencies four levels deep that nobody has looked at since they were written. The FFmpeg flaw Mythos surfaced sat in a single line of code that fuzzers had executed 5 million times. Until recently, the cost of looking deeper was higher than the perceived risk. That math just changed. Anyone running reasoning models against their dependency graph in CI yet, or is everyone still in eval mode? The risk is no longer just the code you wrote. It's the code you pulled in, built on, and shipped with. #CyberSecurity #AppSec #SupplyChainSecurity

4

2 Comments

Sierra partnered with Curative to build their AI voice agent that's creating faster, better experiences for members and providers while freeing up their team to focus on what matters most: building deeper relationships

235

5 Comments

Open-weight models like GLM-5 have quickly become incredibly powerful, rivaling closed, proprietary models like Claude for complex reasoning tasks. But there's a catch: the key to making these models truly agentic isn't just raw text generation—it's tool calling. If an LLM can't output reliable, executable JSON, your agent breaks. Our team at FriendliAI has been obsessing over making tool calling actually work in production. You can easily see the impact of our Unified Tool Call Engine in the OpenRouter data below. While others spike up to 28% error rates, we're keeping it under 2%. 👇

6

You can have customers saying nice things and still not have product market fit. Andy Greenfield, partner at TIA Ventures, spent decades doing qualitative research for major CPG brands. He pioneered a methodology to separate polite responses from visceral pain. Getting different audiences together in panels to understand if products were solving real problems. When they'd do these panels, they'd try to get to one specific question: "If this product were taken away tomorrow, would your life be okay or couldn't you live without it?" That's a very different question than "do you like this product?" Jesse Caesar from First Round Capital explains the difference: "If you want to know what your target is doing or how much, go for quantitative research. But if you want to know why they're doing it, or why they believe what they believe, qualitative research gets you that depth of perspective." Most founders confuse customers being polite with customers who actually care. It’s much easier to do that vs facing hard truths sometimes. You ask someone about your product and they'll say "Yeah, this is super useful." But when you dig deeper with the right questions, you realize they don't actually care that much. I can’t overstate the value of qualitative user research whether investing or building a product myself. Talking to 10 customers with very specific questions to gauge how important the problem really is.  It’s also very helpful to cross reference this with quantitative metrics with tools like PostHog or Mixpanel. “Trust but verify”. The real magic is when you do both of these things in concert. For example, a client you interview may say “I really love using your feature that lets me do XYZ” and they may be very specific about the value it delivers…so it may seem like a totally credible piece of feedback, but then when you look in PostHog or any user analytics platform, you see that they’ve only used the feature once. So this begs the question “if you find feature XYZ so valuable, why have you only used it once in the past month?”  This gap is often the difference between a company that scales and one that plateaus.

20

11 Comments

Our partner, Greenhouse Software, recently released their 2025 Workforce & Hiring Report (https://lnkd.in/gXHd6kdv), and it highlights a challenge we see every day in hiring. "Hiring is stuck in an AI doom loop," says Daniel Chait, CEO and Co-founder of Greenhouse. "A more human and three-dimensional hiring process that helps candidates showcase their skills and focus their job search is the only way to cut through the chaos and connect the right people to the right roles." This message really resonates with our mission Criteria Corp. We believe technology should empower the human side of hiring, not reduce it to algorithms. Our approach—combining proven organizational science (https://lnkd.in/g7qgjn7Z) with a candidate and client experience designed to reveal the whole person—is built to help hiring teams make better choices with more confidence. That’s why we advocate for hiring methods that are personal, skill-focused, and fit-oriented—because that’s the most effective way to match talent with opportunity. Here’s how we help organizations make that vision a reality: ·      Multidimensional assessments (https://lnkd.in/geXe97YJ) & structured interviews (https://lnkd.in/gW268aem) — In as little as five minutes, we surface proven and actionable insights far surpassing resumes or ad hoc interviews. ·      Candidate integrity features — Built-in security measures, clear AI usage guidelines, and authenticity-focused interview structures ensure you’re evaluating the real person behind the application. When candidates are seen and heard across all dimensions, hiring shifts from a resume-matching exercise (i.e. AI doom loop) to unlocking real possibilities. We’re proud to partner with BambooHR and other leading organizations (https://lnkd.in/gyaTSWJ7) committed to making hiring more transparent, equitable, and profoundly human. Let’s build workplaces where people stand out for who they truly are—because the future of hiring is three-dimensional.

20

I like Bret's more positive take on the "saaspocolypse" and how today's software companies can embrace AI to transform. The value in established software companies is the domain knowledge of their customers and industry workflows. Interesting how Sierra is doing outcome-based pricing and not just tokenized consumption. In 30 years we've gone from perpetual licensing, to subscription per user per month, to token consumption and now outcome-based pricing.

11

Most enterprises are using AI wrong. 🤔 I recently sat down with the team at the AI Agents Podcast to talk about a hard truth: A "foot wide, mile deep" approach is the only way AI actually delivers ROI in the enterprise. We don’t need more "chatbots" that give general advice. We need Vertical AI Agents that understand the difference between a financial plan and an allocation strategy. In this episode, we dove into: 🎯 Why horizontal AI fails where vertical AI wins. 🚀 The "Multi-Agent" architecture we’re building at Intelo.ai. 👉 Why the best AI strategy is one that makes humans more creative, not less. If you’re a leader trying to figure out where "Agentic AI" fits into your 2026 roadmap, this conversation is for you. Check it out here: https://lnkd.in/eCTxBbAM #AIAgents #RetailInnovation #DigitalTransformation

23

2 Comments

One pattern keeps repeating across operators adopting AI. Operational AI breaks the moment exceptions aren’t designed in. In real workflows, missing data, vague descriptions, and edge cases aren’t rare, they’re normal. The teams succeeding this year design AI to execute decisions with explicit confidence rules. High confidence? Execute automatically. Low confidence? Route to the right expert before the workflow fails. Practical takeaway: If your AI can’t pause, escalate, and explain its decisions, it’s not production-ready for regulated operations.

7

1 Comment

Love it or hate it, being in the SF Bay Area is still one of the biggest hacks for First Builders to discover huge opportunities before anyone else. But what can you do if you’re not here? Cecilia for example moved here in 2001 and instantly joined Yahoo after feeling the palpable energy around early internet. And it’s been nonstop ever since as she later joined Amazon, Cruise, Replit, and her own breakout company GC AI. She credits a lot to the emerging tech density and optimism in the Bay Area. But she also mentions additional ways people can identify these opportunities early on: - Follow intuition around big waves - Follow your envy - Never lose curiosity - Watch the teen hackers - Talk to peers early and often (builds long term networks/insights)

21

3 Comments

Spotlighting two of our more interesting and meaningful Perpetuate May Portfolio AI Leadership opps. Where Frontier Models meet the complexities of Physical World Systems:

7

More than $4 trillion in equity compensation (RSU, ISO, ESPP etc) remains unmanaged for employees. Founded by veterans in this field, Grantd helps financial advisors transform this complex, often-overlooked benefit into actionable tax-informed strategies for clients and new business opportunities for advisories. We’re excited to see our strategic partners share this vision and place their trust in the team.

8

I was CPO at Blend before their $4 billion IPO. We rejected the “forward deployed engineer” Palantir model that everyone is copying now. Here’s why: Blend could have followed the forward-deployed engineer approach. Both our CEO and CTO were early Palantir. They'd seen it work. Instead, we built a SaaS product with core components. Move fast, iterate, use data to learn. Not custom code for every client. That Palantir model is having a resurgence. Now Sierra, Decagon, and others are hiring forward-deployed engineers. But it’s not the right approach for every company. It’s fundamentally the wrong answer for Assembled’s customers. Here's what we understand about support teams and why we’re not going the forward deployed route: Customer Support teams are used to getting no resources and no support. On the surface, you'd think they'd be excited about engineers building custom solutions. "Finally, someone's offering us developers!" The reality is support teams are terrified of getting help they can't afford to keep. Support teams are wary because they know the math. They don't have budget for expensive, scarce resources. At some point the shoe drops. The engineers disappear when priorities shift. And custom solutions break. Support is left holding the bag with something they can't maintain. They're resource-starved by design. The idea of a dependency on resources they can't control is terrifying. So instead, the goal should be to give support teams the ability to manage, change, and solve their own problems. Not dependencies on engineers they can't keep. Smart, competent people who teach them to do it themselves. Palantir wasn't even a product company for years. They were a custom development shop with reusable components. Show up, build something bespoke, maybe extract learnings. That's not scalable for any but the largest support teams who need sustainability. Forward-deployed engineers are the new shiny object. I wasn’t a fan of it at Blend. And I’m not a fan today either. What I see in the customer support space is that teams need empowerment, not dependency. I want the teams we work with to be able to say: "I love that you’re partnering with me. And I love that I can do this myself too." That's the difference between building a product and building a consulting practice.

80

6 Comments

We're watching a pattern repeat across industries. AI gets deployed. Metrics improve. Leadership celebrates. Then six months later, the system is quietly wound down. What happened? The AI worked. But the organization couldn't absorb it. Downstream systems couldn't handle 10x throughput. Teams routed around it because they didn't trust edge cases. Compliance couldn't audit what the agent actually did. Operations had no playbook for when things broke. Technical success without operational readiness is just expensive theater. If your infrastructure can't handle what AI produces, speed becomes a liability, not an advantage.

2

The best part of being at #Dreamforce is talking to many (MANY!) customers and executives about their biggest challenges in adopting AI agents. The same themes keep surfacing: figuring out where agents actually add value, integrating them into existing workflows, and making sure teams trust the output. My take on it (which has been true for most AI applications for years): 1️⃣ Start with a contained workflow. Don’t “launch an AI agent” . Instead, automate one painful, repetitive process, like summarizing customer calls or drafting support emails. 2️⃣ Focus on clean, connected data before fancy models. Invest in data context. Give your agents access to the right internal docs, CRM notes, and policies and make sure your data is clean, high-quality, and connected. That’s where the real power of agents comes from. 3️⃣ Build trust loops. Let humans review early outputs, collect feedback, and refine prompts and policies before scaling. Focus on the handoff between agents and humans, not replacement. It’s been fascinating to see what’s working (and what’s not) for our customers across many industries. How are you approaching AI agent pilots in your organization? how can we/I help? #salesforce #dreamforce #aileadership #agenticenterprise

112

3 Comments

Incredibly fun conversation with Martin on the a16z podcast. If you're interested in spicy takes on usage-billing, AI value capture, and why many companies are NGMI, encourage you to listen in.

253

21 Comments