Helena Merk

New Uncovered Bookmarks is out. This one covers our Airtree Studios launch (featuring Tim Davis from Modular and a conversation on non-technical operators now shipping code with Saahil Bijlani, Renee Zhang and Jason Zhou), plus some things I've been reading on data centres building their own power plants, Stanford turning papers into AI agents, and why enterprise SaaS might have more moat than markets give it credit for.                                                                                                      https://lnkd.in/g-Xh2pbE

28

1 Comment

Elie Schoppik has spent his career at the intersection of education and engineering. He built early ed-tech tools that helped people actually retain knowledge, founded a Bay Area coding bootcamp that trained thousands of developers through practice and feedback, and now leads Technical Education at Anthropic. Despite all that experience, Elie says most teams are using AI wrong: they’re treating it like an all-knowing oracle instead of a tutor. This is why Elie designs prompts, examples, and workflows that treat AI as a teacher, instead of a shortcut. In our conversation, we discuss: • Why relying on AI for “fast answers” erodes judgment and verification • How to turn prompts into guided lessons • Why context engineering is the most underrated skill in 2025 • Practical exercises any team can run tomorrow to raise AI fluency • What leaders should steal from Anthropic’s approach to technical education  Episode is live now — full conversation linked in the comments.

37

2 Comments

Today, we're announcing our $125M Series A led by Lightspeed with participation from Greylock Partners, Unusual Ventures, Artisanal Ventures, and A*. This round values Resolve AI at $1B. I've spent two decades building production systems and observability tools. In that time, I've learned that operating software is where most engineering time goes: 3AM pages, the cost optimizations to hit margin goals, the laborious compliance questionnaires. Production is the hardest part of software engineering, and every engineer I talk to has lived this story. AI is helping us write more code than ever, but that code still has to be deployed, monitored, and debugged in production. We're scaling code generation exponentially without scaling our ability to operate it. This is the gap Resolve AI closes. We're building AI for prod: agents that work across code, infrastructure, telemetry, and knowledge. Agents that bring together the expertise of multiple teams in every execution, absorbing context and learning continuously. Our goal is to help every engineer operate production fluently. Seeing the impact that customers like DoorDash, Zscaler, and Coinbase achieve with Resolve makes this work worth it. I'm incredibly proud of the team we've assembled to do it. Read my full announcement, linked in the comments

613

151 Comments

At Stand, we build models to understand and mitigate climate risk. Our work depends on an ecosystem of publicly available environmental data, now subject to risk of federal funding cuts. To protect that data and support climate research, we are introducing a Climate Data Commons. This federal data—such as those from NOAA, USGS, the U.S. Forest Service, LANDFIRE—are the backbone of climate risk science. We, along with researchers across the country, are increasingly concerned about its long-term stability and reliability. Internally, we’ve been working to aggregate, clean, and centralize these datasets as part of our internal pipeline. We now hope these will become important tools to support the ecosystem, ensuring the availability of datasets that power billions of dollars of climate research. If you are actively working on research or have been part of recent cuts, and would like to help us keep this data available for the world’s researchers, please reach out to us at datacommons@getstand.com. We’d love to work with you. Please read more in-depth here as we roll this out over the coming months. https://lnkd.in/enpjEPDG

134

5 Comments

Some of my takeaways from an agentic systems roundtable in SF today. Memory isn’t just “store more vectors” Let’s unpack memory into four buckets — core, episodic, semantic, and procedural — and then face the two questions teams usually dodge. First, how do you write information into each bucket. Second, how do you read it back out. Get that plumbing right and the UX sings. ChatGPT feels personal partly because it remembers your late-night rants and procrastination spirals and still picks up the thread. The catch is measurement. We still do not have a solid yardstick for “good memory.” Agent benchmarks are messy already; layer memory on top and we are back to flying blind. Single vs Multi agent Let’s look at Devin versus Claude Code and how the two stack up architecturally. Devin, out of Cognition, sticks to one well-tooled agent running tight feedback loops. Claude Code takes the leader-follower route, farming work out to smaller helpers. That opens the door to task delegation and even lets users spin up their own side-agents, so personalization can run wild. In the end, the architecture you pick hangs on the task at hand and what “done” actually means—and, for the record, multi-agent setups make evals a nightmare to get right. Big takeaway: Nail the evals, then choose the architecture. Predictability beats “looks cool in a demo” every time: especially when an enterprise customer is on the other side of the UX. Thanks to the hosts: Gloria Zhang, Sylvia Tong, Xiaoxiao(Annie) Qin and kevin fang for setting this up.

20

1 Comment

Is Carbon Voice a competitor to meeting notetakers? It’s a great question that I’ve been asked a lot lately. Yes, you can use Voice Memos in Carbon Voice to record a meeting. Voice Memos are automatically transcribed, and AI can summarize, extract action items, or even answer open-ended questions. In that sense, there are similarities. But what truly sets Carbon Voice apart is that you DON'T NEED TO HAVE THE CALL OR MEETING IN THE FIRST PLACE. Here’s how to think about it: 🗓️ Meeting notetakers are for people who love meetings but hate taking notes. 🎙️ Carbon Voice is for people who love collaboration but want outcomes fast. No waiting for calls to get scheduled. No slowing down for yet another Zoom. Just moving at the speed of thought. Every conversations automatically documented, summarized, action items, and turned into action items and shared knowledge. Love meetings? Use a notetaker. Value velocity collaboration? Use Carbon Voice.

10

2 Comments