Kyle Wild

AI has always been about algorithms running in context, where both the algorithms and the context are engineered. AGI is an AI algorithm that is so good it can deal with whatever it encounters, so engineered context is no longer needed. AGI is science fiction. Instead, the world is ramping up on the kind of AI that needs engineered context. And that means many people are about to discover what computer scientists have known for decades: There is no general solution to engineering context because the devil is in the details and there are endless details. It turns out that domain-specific, engineered solutions are the only way to give your AIs the context they need. These solutions are expensive because building and maintaining them is hard. So, buy all the AI you want and push it to the limit. When you realize that the AI doesn't really know your business, doesn't really know your process, doesn't really know how the world works... that's when you call people like me. I'm helping build the context layer for enterprise sales at endgame.io. If you're using AI for GTM, Endgame provides the context your AI needs.

72% of founders say the entrepreneurial journey has hurt their mental health. 81% never talk about it. James Oliver Jr. built the Kabila Founder Mental Health Fund to do something about it - direct therapy grants for founders who can't afford to pay for it themselves. No complicated process. No strings. For the month of May, I'm matching every dollar donated, up to $25,000. Your $150 becomes $300 - four therapy sessions for a founder who needs them. I also wrote Chapter 4 of James's new book, Burn Bright, Not Out, and I'm giving the chapter away for free. If it resonates, buy the book - 90% of net profits flow back to the fund. https://lnkd.in/gfP59NrS

72% of founders say the entrepreneurial journey has hurt their mental health. 81% never talk about it. James Oliver Jr. built the Kabila Founder Mental Health Fund to do something about it - direct therapy grants for founders who can't afford to pay for it themselves. No complicated process. No strings. For the month of May, I'm matching every dollar donated, up to $25,000. Your $150 becomes $300 - four therapy sessions for a founder who needs them. I also wrote Chapter 4 of James's new book, Burn Bright, Not Out, and I'm giving the chapter away for free. If it resonates, buy the book - 90% of net profits flow back to the fund. https://lnkd.in/gfP59NrS

If you're in GTM, how many times has "our team will be slow to adopt that" come up in a meeting about a new tool? Keep your hands raised if that same conversation came up again in the next three follow-up meetings. That line has derailed more software rollouts than bad product ever has. And almost every time, it's really about the UI. Too many clicks. Doesn't fit the workflow. Reps won't log in. The tool is fine, getting humans to use it correctly is the actual hard problem. Salesforce shipped something Friday that I think quietly makes that problem obsolete. It's called Headless 360. Agents can now do everything inside Salesforce without a human ever opening the browser. No UI required. The data and the logic are fully accessible programmatically, which means the interface your reps have been slow to adopt for years just became optional. And that's worth sitting with for a second, because it runs deeper than a product update. Enterprise software for the last 25 years has been built around one core assumption: humans navigate interfaces. Every design decision, every onboarding program, every CRM adoption initiative your RevOps team has ever run was built on that assumption. The whole business of enterprise SaaS, the implementation partners, the admin certifications, the change management playbooks, exists because getting humans to use the interface correctly was the hard problem. Agents don't have adoption problems. They don't complain about the number of clicks. They don't need three rounds of enablement. They just consume the data and do the work. Which means the new hard problem is what they're consuming. An agent pulling from a stale deck, an outdated pricing doc, or a CRM field nobody's touched in six months doesn't fail loudly. It just confidently gives you the wrong answer. And unlike a rep who might gut-check something before sending it to a prospect, the agent won't. It trusts whatever it's pointed at. That changes how you have to think about evaluating software going forward. The criteria used to be ease of use, feature depth, and whether it integrated with your existing stack. I think it's about to become, "how clean is the data model, how good are the APIs, and how well does this serve agents consuming it programmatically rather than humans clicking through it?" The vendors who built their moat on having the best interface are going to feel this shift. The ones who invested in the data layer underneath are going to pull ahead in ways that aren't obvious yet but will be very obvious in two or three years. Salesforce doesn't make a bet like this unless they know the game is changing. The question is whether the rest of your stack is ready for the same shift. My CEO Alex Bilmes wrote about what happens when the data layer breaks down at scale. If this post has resonated, it's worth a read. Link in the comments.

In the last few weeks, I've saved our engineering team about 25 hours, caught six production issues before they impacted customers, killed 92 percent of pointless error noise and made our systems run smoother and smarter with this one simple trick. You're probably wondering, wow, Andy, is this written in such a way to make me keep reading? Yeah. It totally is. Nice job, LinkedIn reader. So, previously, our engineers would sometimes look at logs in our various systems occasionally—some would check in the morning, some in the evening, some would just go "I wonder if I can find a problem or two." It wasn't any kind of regularly coordinated thing. And then I thought, we should come up with a process for this because they find issues, my brain going into manager bureaucracy mode, thinking who could be on call, how we could rotate it, etc. But instead, I had a flash of pure genius: I'll just automate this shit. Now, every morning before I even go to my computer, Claude Code automatically checks our production infrastructure — error logs, Argo workflows, Temporal health, Sentry, deployments, and more — and triages findings by severity, and posts a summary to Slack. When it finds real issues, it investigates root causes and opens draft PRs with fixes. I then test the work myself before sending it off to our engineers to take a look and get it out the door. Has it found false positives? A few times. Has it prevented issues before they began? You bet your ass. Now it runs in the afternoon, too, providing insight into work we've deployed throughout the day. My little morning DevOps Bot also keeps track of the last few weeks' worth of issues in a markdown file, which it updates after every run, making sure we're continuing to keep a track of issues and proving we've actually fixed issues over time. What do you have to learn from this? Instead of adding more dreaded bureaucracy, automate your shit. Feel free to tell me if you disagree below. You'll be wrong, but, live your truth (incorrectly).