CTO gathering in San Francisco - Pebblebed VC.

We are bringing together 40 to 70 technical leaders and CTOs from Bay Area companies for an informal evening of networking, short talks, and open discussion around AI and the hardest technical team challenges. Request Your Seat.

Tinybird is sponsoring the event so far. If your company wants to sponsor this San Francisco gathering too, reply to this email.

At our last CTO Circle gathering in Barcelona, the most voted question during the open discussion came from an engineering leader at Spotify:

“How do you keep improving as a technical leader?”

I loved the first reply, from Andrea at Microsoft: asking the question is already the first step to solving it.

Most technical leaders are so busy with fires, delivery pressure, hiring, stakeholders, and team issues that they rarely stop to think about improving their own craft.

Then the main answer from the room was simple: find people who are a few steps ahead of you, and learn from them.

Not necessarily a formal mentor. Sometimes it is just someone whose work and career you admire. Someone who has already faced the kind of problems you are facing now, and can give you honest perspective.

Personally, that has been huge for me while building Pullpo. I went from being a CS student with almost no idea about companies, sales, enterprise customers, fundraising, or product strategy, to figuring many of those things out through conversations with people who had already been there.

One important lesson: you usually do not need to give equity, pay expensive advisors, or join a mentor program to get help.

When I started my company, some advisors and agencies offered mentoring in exchange for equity. My advice today would be: be careful with that.

Most good mentors help because you are genuinely trying, because you come prepared, and because you ask about a real problem with enough context.

A bad way to ask is: “Can you be my mentor?”

A much better way is: “Hey Josh! I’m [intro]. I’ve been following your work at [background] and really admire it. I’m struggling with [problem that has to do with background]. These are the options I’m considering, and this is where I’m stuck. Would love your perspective.”

That changes everything.

A few practical ways to keep growing as a technical leader:

1. Write down one leadership problem you are currently facing before asking for advice. “I want to become a better leader” is too broad. “My team is missing deadlines because priorities keep changing” is useful.

2. Build a small group of people you can ask for honest feedback. Ideally, some are ahead of you, some are at a similar stage, and some work in different types of companies. You can also ask at #ask-anything and meet new leaders bi-weekly in #virtual-coffee-1-1 or at our in-person events.

3. Don’t look for one perfect mentor. Look for different people for different problems: hiring, architecture, communication, managing up, product strategy, fundraising, and more.

4. After every useful conversation, write down one thing you will actually try. Otherwise, mentoring becomes just another interesting chat.

5. Review your own leadership every few months: what problems keep repeating, what conversations you are avoiding, and what skill you need for the next stage of your team.

Anthropic launched Claude Sonnet 5, positioning it as its most agentic Sonnet model yet: closer to Opus-level coding and tool use, but at a much lower price point and now available across Claude plans and Claude Code.

Anthropic is redeploying Fable 5 after export controls were lifted on June 30. Global access resumes on July 1, and the broader lesson is that the industry still needs a shared standard for judging jailbreak severity and model risk.

Qwen 3.6 27B is the sweet spot for local development. The case here is simple: it may be the first local model that feels broadly useful for real development work, even if it runs hot and slower than the MoE variant.

Software is becoming marketing. If software gets cheap to build and easier for outsiders to judge, more of the advantage moves away from pure product construction and toward positioning, distribution, and services.

AWS is investing $1B in forward deployed engineers, embedding engineers directly with customers to help them ship agentic AI systems faster. Cloud vendors are moving closer to the Palantir, OpenAI, and Anthropic playbook.

Anonymous GitHub account mass-dropping undisclosed 0-days. The repository aggregates public exploit PoCs and unreported vulnerability research, which makes it catnip for security researchers and a headache for vendors and defenders.

Software engineering in the age of AI. Another useful reminder that the job is moving away from just typing code and toward framing problems well, reviewing generated output, and owning reliability, judgment, and taste.

New here? Join the Slack community and introduce yourself.