Tyler Hayden

launching today @TolmoHQ → $22M from Accel & Y Combinator → AI security agents for production environments → Every PR, deploy & alert covered — with the fix attached → 40+ design partners live today

@pbetouin @TolmoHQ Security agents hacking a knowledge graph of production 24/7 - why not?

Let's go! We're out of stealth with @TolmoHQ! AI is completely redefining how software gets built. And how it's being attacked. Helping define what security means for this new world was impossible to resist. We had to go back! We’re building the AI agent fleet to secure production environments. Our security agents continuously work on a live production knowledge graph that connects data from code, cloud services, containers, databases, and third-party providers. Imagine high fidelity pictures of what you’re building, with a tracked history of changes at scale! Our team spent years on the offensive side, breaking into cloud services, kernels, or cryptographic implementations, as well as building observability and security solutions for thousands of customers in production at Datadog. Over the last few months, we’ve raised $22M, led by Accel and Y Combinator, assembled a stellar team in San Francisco and have been iterating with our first 45 design partners. It’s go time! Joining me today to announce Tolmo: my cofounders @JbAviat, @poledesfetes, and @arnaud_breton, alongside @msuiche and our early team @wbeuil, @jimmycaputo, @cryptonitemmk, @theraniarnia, and @TylerHaydenSF. We're psyched to embark on this new journey together! We're building Tolmo in the open, and we're looking for feedback. Would love to hear your thoughts about what we're building! tolmo.com/blog/introduci…

Cursor Composer 2.5 looks competitive. Cost per task is interesting metric not captured by benchmarks

I Encouraged a Designer To Accept 30 Terminal Commands She Didn't Understand... On a Public Company Machine linkedin.com/pulse/i-encour…

I got demo'd by an AI Sales platform startup today. Asking for $25k with no trial on the first call is a bold GTM move. Their pitch: "We have so much demand we can't offer POCs... annual commitment only." That's the FOMO close. It only works when your product can carry the weight and they gave no evidence of that. The testimonials on the website are most likely investors, friends of founders - you know they roll deep - or free beta users.

@sebuzdugan big things happening. constantly surprised how people are NOT using these tools

@TylerHaydenSF Big claim, let’s see 😅

"Claude and codex will probably wipe out half the tech industry" - engineering friend

Quick catch up if this is new: Agent Skills were popularized by Anthropic. They were introduced as a feature to make Claude more capable with better context in Oct 2025. They help agents do things. They became an open standard in Dec 2025 and the standard is adopted by OpenAI and others. The standard is here: agentskills.io/home

Agent skills are just context injection. But they're also how you become indispensable in the workforce and to the AI agents building with your tool.

If your SaaS product isn't in the agent toolkit, you don't exist. Agent-augmented work is happening in 2026. The skill databases are where discovery happens.

Your documentation is already 80% of a skill. You're probably leaving traffic on the table by not structuring it for agents. Agents don't context-switch to ChatGPT. They use what's in their harness toolkit (Skills!).

Agent skills are how you get discovered in an agent-driven world. Well-documented, opinionated skills = locked-in users

@addyosmani @github Claude Code has made this insanely easy as well. Beyond the subagents feature, you can just prompt it to “create a some 4.5 agent for each task/file/etc in parallel to preserve the context window of the primary orchestrator”. It’ll create and run 4 parallel agents at a time

The next big thing is Asynchronous Coding Agents. Such Agents change our role as engineers from Conductor (directing one agent at a time) to Orchestrator (defining tasks for a "fleet" of agents working in parallel), for very well-defined tasks with human review. @github's Copilot Agent, as covered at #githubuniverse, is one such example I've been trying and liking. There are a number of options also worth considering, including @julesagent (also great!), Codex web by @OpenAIDevs, and Anthropic's Claude Code for web. Editors like @cursor_ai are also exploring this with some very interesting UX ideas. For the more advanced, orchestrators like Claude Squad, Magnet and Conductor by @charlieholtz (nice balance on the engineering/product spectrum) are certainly worth considering. Some caveats: Orchestration has been working well for me in small-medium sized projects: for fixing UI bugs, adding small features etc. This can be good for tackling migrations, upgrading dependencies etc. Enterprise teams are very much still figuring out their "Conductor" patterns but eventually, I do see more orchestration becoming a standard part of the engineering workflow.

@poledesfetes Try Pulumi 😏

When I finally start to think that this "code in yaml" pattern could work, I see a random limitation in Github actions format and I remember how stupid it is.

Introducing Pulumi ESC: Environments, Secrets, and Configuration - A new solution to manage application and infrastructure environments, secrets, and configurations. 🧵 pulumi.com/blog/environme…

Can't wait!

If you're in Seattle next Tuesday, join Team Pulumi to celebrate our Series C and biggest week of product launches ever 🚀 The more the merrier. And hey, the beer and BBQ are on us! 🍻 meetup.com/pulumi-seattle…

Congrats to all at @PulumiCorp ☁️☁️☁️

@zaid_ajaj @BicepLang @PulumiCorp Congrats Zaid, nice write up

Heyy folks, excited to share with you the project I've been working on: a converter plugin that extends the Pulumi CLI to convert @BicepLang code to any @PulumiCorp supported language. It is written in #fsharp of course 💜 learn more about it here: pulumi.com/blog/convertin…