Ben Go 🌎

Building in SF + AI right now feels like a privilege. You're not just at the forefront of Technology - you’re surrounded by technical builders who live and breathe their craft. People who’ve gone deep in their niches and are now rethinking the fundamental workflows that power the world. Had a deep dive yesterday with Shreya Karpoor from Simplex on long-context browser agents - we have two completely different use cases, but similar problem-solving frameworks. It’s rare to be able to talk about tradeoffs in context chunking, tool switching, or browser orchestration without having to explain what those words even mean. And it’s not just theory. You see someone jump from Cursor to Claude Code in real time, try it out, and you’re converted in 10 minutes. That’s the bar. That’s the pace. And now, there's 5 more folks building a better Claude Code too! Everyone here shares the same hunch: AI agents are about to transform every business-critical flow. And being part of building that future, side by side with others who are just as obsessed - what a time.

143

7 Comments

I've joked in the past that what *really* makes LLMs work is our tendency to see faces on toast, but there's a more serious point there about how much of our perception of the ability of models to "understand", "reason", "follow instructions" etc is in reality projection. We've evolved to read intention into the behaviour of other people so that we can predict what they might do. But we can also see intent in the behaviour of pets, weather, dishwashers, etc etc. So we shouldn't be too surprised if something that's designed to statistically reproduce human creativity and reasoning has that effect on many of us to a much greater extent. I certainly fell for it during the first few hours experimenting with GPT-4, until I played it at chess, and then the curtain was pulled back. It doesn't know where the pieces are on the board, it doesn't plan ahead, it doesn't know the rules. It literally just predicts - by matching the sequence of moves to its vast example space of chess transcripts - which chess move is most likely to come next. Once you've seen it, it can't be unseen. But I appreciate that a lot of people have yet to see the tiger in the Magic Eye picture. The dazzling complexity of human language makes it hard to see the wood for the trees. That's why something simple and deterministic, like chess, makes it much clearer. As impressive as LLMs can be, I encourage users not to mistake powerful pattern matching and next-token prediction for actual intelligence or understanding. I urge folks who use these tools - which is all they are - to take a rational and evidence-based approach to them, as I've been doing for 2 1/2 years now. Your cat doesn't understand what you're saying. It can learn to recognise certain words, your tone of voice, your body language, and associate it with - for example - imminent treats or bath time. That learned behaviour can be easily mistaken for actual conceptual understanding. Clever Hans couldn't do arithmetic when he couldn't see his trainer, but not even his trainer realised he was subconsciously giving off visual cues. Oh yeah, and LLMs don't understand the instructions and rules in your claude.md file. (A good test is to add a "brown M&Ms" rule to your context.) But we're hardwired to see that in them, and it's a very powerful effect. I see much confirmation bias, for example, in interpreting output - a strong desire to focus on the things they get right while overlooking many of the things they got wrong. And they get a *lot* wrong. Expect that, because it's not going to get much better. You have to keep these tools on a very tight leash.

46

12 Comments

At Trustible, we are AI Pragmatists. Not Doomers. Not Boomers. But, a federal moratorium on state & local AI laws would’ve done more harm than good for the innovation agenda it argued it was trying to protect. Trustible applauds the Senate’s decision to strip down the AI Moratorium from the reconciliation bill – though we expect this conversation to continue to evolve. Our strong perspective is that completely banning local & federal AI regulation would’ve slowed down AI adoption, not accelerated it. Why? ◾ It would’ve eroded public trust, not just through bad actors—but through the perception that AI is unregulated and unaccountable. ◾ It would’ve pushed liability onto startups and smaller players who can least afford the risk. ◾ It would’ve created legal limbo and litigation risk, freezing procurement and delaying deployment. ◾ It was solving a problem that doesn’t exist—most bills mentioning artificial intelligence are not actually regulating AI. ◾ It would’ve stifled the kind of information sharing we need to improve safety—especially for those without in-house research teams. Our customers (who range from enterprises, startups, non-profits, and government institutions) are telling us that they are looking for real-world frameworks that help them innovate responsibly—and earn the trust of those who matter most to them. Read our full perspective in the blog post in comments 👇

33

2 Comments

If you're like me*, you've been trying to get to the bottom of what the term "Forward-Deployed Engineer" (FDE) really means in agentic startups. I found this YC podcast (below) with Bob McGrew (former Chief Research Officer at OpenAI and early Palantir leader) to be super actionable. Some things that I took away: * For domains where it's not yet clear there is a one-size-fits-all product, FDEs make sense. * The Palantir FDE team had two roles - someone with customer domain experience (called "echos") and more technical role (called "deltas"). * A big part of the approach is to prototype rapidly, even if it felt initially like a one-off. * That being said, Palantir's core product team still had to build a general purpose platform that evolved as the FDE teams learned. * Being an FDE is a powerful onramp for aspiring founders. I learned a lot! *Yes being "like me" isn't recommended 😭 https://lnkd.in/gV3ejNqn

52

16 Comments

MIT just published a staggering stat: 𝟵𝟱% 𝗼𝗳 𝗲𝗻𝘁𝗲𝗿𝗽𝗿𝗶𝘀𝗲 𝗚𝗲𝗻𝗔𝗜 𝗽𝗿𝗼𝗷𝗲𝗰𝘁𝘀 𝗳𝗮𝗶𝗹 𝘁𝗼 𝗱𝗲𝗹𝗶𝘃𝗲𝗿 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝗶𝗺𝗽𝗮𝗰𝘁. Why so many fail: - Tech-first pilots that don’t map to outcomes. - Governance blind spots — underestimating policy, security, and what it takes to be enterprise production-grade. - Proof-of-concept purgatory — slick demos that never go live, a graveyard of gimmick projects. - Operational reality — work is messy and cross-departmental, and most projects ignore it. - Orchestration failure — even when pilots make it to production, they remain siloed and impossible to connect or scale. 👉 The lesson: it’s not about more models. It’s about building a system of action. That’s why we built ai.work. I couldn’t stand watching great companies waste time and money on failed SaaS projects and now it's happening with AI! We deploy AI Workers that: ⚡ Operate inside the flow of work (event-driven, not chat-driven). ⚡ Are enterprise production-grade from day one with policies, approvals, security, and auditability. ⚡ Understand organizational context and orchestrate multiple agents across IT, HR, Finance, and Ops. ⚡ Plug directly into Salesforce, Jira, ServiceNow, Workday — just like a real employee would. ⚡ Are measured on outcomes that matter: MTTR, auto-resolution rate, cost-to-serve, productivity. The Industrial Revolution wasn’t about pilots sitting on the side. It was about machines doing the work. The AI Revolution will be no different. 95% will fail. The 5% will redefine work. Join the 5%.

66

8 Comments