Good questions

I once sat in on a board meeting at a company where Peter Fenton was a board member. I only saw him in action once, but I've thought about it ever since.

The CEO shared an overview of the business, and various VPs presented updates about their functional areas. Over an hour passed, and Fenton just sat there and listened. Then he asked one question. I can't for the life of me remember what it was. But it completely changed the direction of the conversation for the rest of the meeting. We had been layering on cruft for an hour and he cut through all of it in a single sentence.

If I could master anything, it would be asking questions like that.

The best framework I've found for understanding the art and science of asking good questions comes from an unlikely place: a book called Ignorance, by Stuart Firestein, a neuroscientist at Columbia. His argument is counterintuitive. We tend to think ignorance precedes knowledge: you don't know something, so you go learn it. But in science, it's the other way around. Knowledge precedes ignorance. The more you learn, the more precisely you can identify what you don't know. He calls this "knowledgeable ignorance."

That's the best explanation I’ve seen for what Fenton was doing. His question wasn't the product of confusion. And it wasn't a performance, either. It was the result of decades of pattern matching across hundreds of companies, which had given him the ability to identify, very precisely, the one thing we weren't talking about. His ignorance was informed. It was the product of expertise, not the absence of it.

But identifying the gap is only half of it. You also have to construct a question that moves the conversation somewhere useful. I saw this happen recently. A company was presenting stellar month-over-month revenue growth. Everyone was nodding along. The obvious move was to talk about how to accelerate. Then one of the board members asked what sounded like a dumb question: "why are we growing?"

Well, it turned out the VP Sales didn't have a great answer for why certain deals were closing and others weren't. And, indeed, there was reason to think the good numbers might be masking an issue that could show up down the line. None of that was in the board deck, of course. And it only surfaced because someone asked a question so basic it felt almost rude.

That's what a great question does. It reframes the conversation in a way that promotes honest reflection. "Why are we growing?" isn't a “gotcha” question. It’s not a trap. And it isn’t trying to force some confession. It's simply an invitation to think out loud.

The best questions also almost never sound smart. They sound obvious, or naïve, or uncomfortably simple. And that is exactly why most people don't ask them. Instead people monologue, imagining that at least then they'll sound smart. Or they lead the witness, framing things to get the answer they want to hear.

That’s the paradox: the best questions are usually only available to people who know a great deal, and yet they often sound like they came from someone who knows almost nothing.

An hour of decks and updates and performance can take a meeting in circles. Sometimes the most valuable person in the room is the one who says almost nothing, and then asks the question that makes everything else look beside the point.

David Peterson

Could the future of software be fluid
How do we get the best of AI without losing the soul of software?

The future belongs to young missionary teams
Why it makes more sense betting on youth in the current moment

The AI-native enterprise playbook
Ten real-time observations on a rapidly evolving playing field

No more painting by numbers
It’s the end of the “SaaS playbook.

ENTERPRISE/TECH NEWS

Lunar Infrastructure - NASA’s Artemis II mission, the first crewed lunar flight in over 50 years, is positioned as a testbed for the next phase of space infrastructure. Beyond sending humans further from Earth than ever before, the mission is designed to validate systems required for sustained deep-space operations, including life support, navigation and long-duration crew performance. It forms part of a broader programme to establish a sustained human presence beyond Earth. Future Artemis missions will rely heavily on private-sector partners — including companies providing landers, spacesuits, and transport systems — marking a shift from state-led exploration to a hybrid public–private model.

Code Exposure - Anthropic disclosed that internal source code for its Claude Code tool was accidentally released during a software update due to “human error”, exposing nearly 2,000 files and around 500,000 lines of code on GitHub. The company stated that no customer data or model weights were compromised, attributing the incident to a packaging error rather than a breach. The material included elements of internal architecture and product features, reportedly attracting more than 29 million views within hours. A rewritten version of the code quickly became one of GitHub’s fastest-downloaded repositories. Observers identified features including an always-on AI agent and a Tamagotchi-style coding assistant. This is the second disclosure involving Anthropic in recent weeks, increasing scrutiny on internal security practices.

Workforce Reset - US tech layoffs have reached their highest year-to-date level since 2023, with more than 50,000 job cuts recorded in Q1 2026, according to Challenger, Gray & Christmas. Companies are reallocating budgets toward AI investment, particularly across engineering and operational functions. Layoffs are concentrated in roles where tasks are being automated or reprioritised.

HOW TO STARTUP

Four Archetypes - Yoni Rechtman argues that the structure of work inside tech companies is shifting away from traditional functions toward a smaller set of working styles. Rather than roles defined by output (product, design, engineering), organisations are increasingly organised around how work gets done. He outlines four archetypes: 1) high-velocity generalists (“product engineers”) who ship across functions; 2) infrastructure and reliability specialists; 3) operators who impose discipline and governance; and 4) and customer-facing roles that translate complexity into accessible experiences. These are not job titles but modes of operating that cut across roles. As tools reduce the cost of producing code, designs, and specifications, organising teams around outputs becomes less relevant. “The right question is how do you approach your job irrespective of the title.” The result is a shift toward multi-functional operators, with engineering, sales, and operations increasingly converging around product and customer outcomes.

HOW TO VENTURE

Capital Overflow - Nicolas Colin outlines that Europe’s relative weakness in venture is not explained by talent, culture or regulation, but by differences in capital allocation. He describes an “overflow mechanism” driven by the dollar’s reserve currency status. Foreign capital flows into US safe assets - Treasuries and large-cap equities, saturating the low-risk end of the market. This displaces domestic institutional investors, pushing them up the risk curve into venture capital. Venture in the US is, in part, a byproduct of excess global savings being absorbed and redistributed through financial markets. A second factor is historical. The US ecosystem that absorbed this capital was built through decades of state spending tied to wartime and geopolitical urgency. Institutions such as Stanford redirected public funding into commercial innovation, contributing to the emergence of Silicon Valley. Outside the US, this mechanism does not exist in the same form. Venture funding is more reliant on the state and large private fortunes, with public capital accounting for a significant share of European investment.

Velocity Funding - Recent AI fundraising rounds are compressing both timelines and underwriting standards. In late 2025, DualEntry, an AI ERP startup less than a year old, raised a $90 million Series A at a $415 million valuation, with reported ARR around $400,000. The pattern is not isolated. Series B rounds are increasingly closing within 27 to 60 days of Series As, often without new performance data. Companies including Rillet and Campfire AI have raised successive rounds within weeks at significantly higher valuations.Capital is being deployed ahead of evidence, with progression between rounds driven by momentum rather than updated fundamentals.

Layer, Not Vertical - Lexi Novitske frames a shift in how she approaches AI investing: treating AI not as a category, but as a layer applied across businesses. The focus shifts from standalone AI companies to those that control distribution, operate within systems where money flows, and accumulate proprietary data over time. Her framework prioritises three characteristics: ownership of customer trust and access, embedding within core workflows, and data that compounds efficiency.  She highlights emerging markets, particularly Africa, where companies often operate end-to-end processes, build region-specific datasets, and establish trust in environments where it is limited.The approach reframes the question from which AI companies to back to which problems are being overlooked to which problems are being overlooked, and which businesses are sufficiently embedded that AI enhances rather than replaces them.

FalkorDB
Core Developer (Tel Aviv, Hybrid)

Firebolt
Senior Software Engineer (Munich)

Motorica
Platform Engineer (Stockholm, Hybrid)