Yuval Passov

People mostly start on their own rather than because an employer mandated it, they say, and that is the point: the person who experiments independently is the one who stays relevant.

• The method is deliberately small and repeatable: pick one tool, give it one hour, and see what it actually does for your workflow.
• In an earlier post, they call the same pattern the AI Accelerator Loop: play with tools, understand how each one changes your specific work, then teach it to your group, which they describe as the fastest way to master it yourself.

Passov reports on Marc Zao-Sanders' 2026 study of how people use AI, drawn from 12,637 real examples, and highlights two findings: therapy and companionship ranked first, ahead of coding and writing, and almost nobody adopted AI at work because their company told them to.

• The study calls this shadow usage; the illustration is someone quietly closing tickets twice as fast.
• The suggested response is one tool, one hour.
• Passov closes the issue with a task: build one small thing on a personal laptop.

When Passov marked 13 years at Google, the advice was a structured loop of play, understand, and share. By the later AI-use-research issue, it had compressed into a one-tool, one-hour weekly habit.

Read original · Original 2

They frame this as the change every founder needs to make: from passive responding to active management, treating AI as an extra worker rather than a waiting room.

• The practical question is who inside a company owns the agents doing the work, since Passov expects that role to compound.
• They extend the same idea from software to business models: from charging per seat to charging for a finished outcome.

Drawing on Google Cloud's published collection of 1,302 real-world generative AI use cases, which began with 101, Passov picks out five trends. Among them: a supply chain agent detects a delay, contacts a compliance agent, and triggers a financial agent to adjust a forecast with no human routing the work; a manager asks a 40-year-old SAP system a question in plain English and gets an instant answer without migration or a ticket; and in shipping, an assistant drafts a delay email for review while an agent checks the weather, finds alternative suppliers, negotiates prices, places the order, and reports that the issue is resolved.

Read original · Original 2

Experience alone does not protect anyone, they argue; the security worth having is the ability to change.

• They tell readers to write their own development plan rather than one for a manager, help people without keeping score, and treat flexibility as the only real security.
• They also warn that feeling like an expert is a danger sign, and deliberately take outside courses to escape what they call the internal bubble.

Announcing a 13th year at Google, which Passov calls a Google Bar Mitzvah, they point out that the same job has not lasted that long: they relocated to Singapore and back, switched verticals from gaming to startups, and reinvented the role repeatedly.

• They also treat writing on LinkedIn as a mirror, using it to turn the day's noise into structured, owned knowledge rather than audience-building for its own sake.
• The line "experience won't protect us, but agility will" reappears in a later post about agents.

Read original · Original 2

Passov argues that attention is limited and something has to give, so the goal is to choose deliberately which area to turn down rather than make a failed attempt at balance. They name guilt as the most dangerous background noise for a founder and favour small repeated habits over large plans—the tasks wait for a stronger version of you.

Passov relays Udi Ledergor's four-burner stove analogy: health, family, career, and social life are the burners, the gas is limited, and blasting all four at once explodes the stove.

• Ledergor himself turned his career burner down for two years when his twins were born, without guilt—an example of intentional imbalance.
• Passov applies the same reasoning to personal routines: steps up more than 14% from walking the children to school and turning calls into walk-and-talks; a 30-minute run at 6am in Las Vegas despite jet lag; and a living-room Tabata workout with the kids on a day when sirens, closed schools, and a cancelled gym pushed work aside.

Read original · Original 2 · Original 3

Instead, the job is to build memory links so people think of you when they are ready, roughly 12 to 18 months out—being long-term greedy.

• The three levels of content Ledergor names are exclusive insights from proprietary data, original surveys, and polarizing opinions.
• Passov credits the framework to the guest and uses the same issue to argue that businesses are moving from selling software to selling completed outcomes.

Ledergor's line, as Passov relays it, is that around 95% of a market is not in a buying cycle at any given moment, and that if you are not making anyone uncomfortable, you are probably not exciting anyone either. The outcome-versus-tool distinction comes through in a delayed shipment: an assistant drafts an update email for a human to send, while an agent checks conditions, finds suppliers, negotiates, orders, and simply reports that the issue is resolved.

Read original

They keep returning to where value sits when everyone has the same models: not in the model itself but in the loop between proprietary data, a shipped product, and the next training cycle, and in the messy, specific problems around the model where a generic wrapper competes directly with the labs. They also argue that a harder go-to-market can be the moat, which is why unglamorous sectors get flagged.

Passov covers Y Combinator's Summer 2026 Requests for Startups and picks out AI for government, modern metal mills, and AI guidance for physical work, noting that the tougher sales motion is what makes them defensible.

• They cite a16z's Joe Schmidt on the Yellow Brick Road—the value is in the application layer, proprietary data, human approval steps, and logistics, not in better code or image generation—and Mario Gabriele's Saplings, a study of 260 founders across 560 books.
• From the first Founder Private Forum at Google Cloud Next, which Passov helped organize and where Jeff Dean joined, they take two points: the moat is the data-product-training loop, and founders now plan on two-month rather than six-month cycles.

Read original · Original 2 · Original 3

In gaming, that means whether the game holds people; in services, it means delivering a result rather than a licence. These positions come largely through named practitioners and lists, with Passov adding their own framing.

From a conversation with Daniel Mironov, a former Google colleague now an investor at Vgames, Passov reports that starting a gaming studio is much harder when AI runs through design, code, and art: teams are smaller, burn is close to half of what it was, and everyone has the same AI tools, so the edge returns to gameplay that people cannot put down. They make the same point about services, pointing to AI-native agencies and AI-powered accounting or insurance that sell the outcome rather than the software.

Read original · Original 2 · Original 3

Start here for the clearest picture of the shift a B2B software audience is trying to make.

Passov describes the shift from opening a tool, typing a prompt and repeating the process to managing agents that work without being asked.

• They ground that shift in Google Cloud's published list of 1,302 real-world generative AI use cases, which began with 101, and identify five trends: agents replacing workflows (a supply chain agent spots a delay, contacts a compliance agent and triggers a financial agent to adjust a forecast, with no human routing the work); natural language as the new IT department (a manager asks a 40-year-old SAP system a question in plain English and gets an instant answer without migration or a ticket); media as a factory (one hypothesis turned into thousands of personalized video variations in hours); AI seeing the physical world (a factory floor camera spots a worker without a helmet and sends an alert automatically); and cybersecurity moving from alerts to action (an agent writes a detection rule, isolates a server and deploys a trap).
• They also describe organizing the first Founder Private Forum at Google Cloud Next '26 with Jeff Dean and report two takeaways from the room: the moat is the loop between your data, your shipped product and your next training cycle, and founders are planning on two-month cycles rather than six.

For brands planning a campaign

A strong anchor for a webinar or article on agent management. The five trends provide a ready-made structure, and Passov's question—who in your company manages the agents doing this work?—can frame a practical discussion with an operations or product audience.

A newcomer gets the research finding that explains why bottom-up experimentation matters more than top-down mandates.

Passov reports on Marc Zao-Sanders' 2026 edition of AI in the Wild, a study of 12,637 real examples of how people use AI.

• Two findings stand out: the number one use is therapy and companionship, ahead of coding and writing; and at work, almost nobody picked up AI because their company told them to.
• They started on their own, often without telling anyone, a pattern the study calls shadow usage.
• One example is a person quietly closing tickets for a movie twice as fast.
• Passov frames this as a second job everyone now has alongside their real job: keeping up with AI by trying tools, building small things on a personal laptop and seeing what is real.
• Their suggested starting point is modest: pick one tool and give it one hour.
• The same post also covers a16z's Joe Schmidt on the Yellow Brick Road, the argument that value is in messy, specific problems around the model rather than generic wrappers; Anthropic's confidential IPO filing after a raise at a $965 billion valuation; and Mario Gabriele's Saplings study of 260 founders and 560 books.

For brands planning a campaign

Useful for a message about bottom-up AI adoption. An article or workshop could pair the shadow-usage finding with Passov's one-tool, one-hour exercise, giving teams a low-friction way to surface and share experiments already happening quietly.

Start here for advice on content that earns attention and a plain example of the shift from software to outcomes.

Passov interviews Udi Ledergor, chief evangelist at Gong, who argues that 95% of a market is not in a buying cycle at any given time, so pitching those buyers now pushes them away.

• Instead, build memory links so people think of you when they are ready to buy in 12 to 18 months.
• Ledergor names three levels of content that cut through the noise: exclusive insights from proprietary data, original surveys and polarizing opinions.
• As they put it, if you are not annoying anyone, you are probably not exciting anyone either.
• Passov also relays the four burner stove analogy for founder life: health, family, career and social life are the four burners, but the gas is limited, so at least one must be turned down.
• Ledergor turned the career burner down for two years when their twins were born, without guilt—a choice Passov frames as intentional imbalance rather than failed balance.
• The post closes with a shift from SaaS to outcomes as a service: an assistant drafts an update email for a delayed shipment, while an agent checks the weather, finds alternative suppliers, negotiates prices, places the order and simply reports that the issue is resolved.

For brands planning a campaign

Directly relevant to AI adoption thought leadership that has to compete in a saturated content market. Ledergor's three content levels can shape an editorial plan, while the outcomes-as-a-service example gives a buyer-facing way to explain why customers now expect results rather than seats or tools.

A newcomer gets a personal, repeatable method for learning AI tools rather than waiting for formal training.

Passov marks 13 years at Google and pushes back on the idea that staying that long means settling.

• They describe moving to Singapore and back, switching verticals from gaming to startups and reinventing their role.
• Their AI Accelerator Loop has three steps: play by testing new tools, understand how a tool improves your specific workflow, and share by teaching it to your group, which they call the fastest way to master it yourself.
• They also treat LinkedIn writing as a mirror that structures vague ideas into owned knowledge and argue that feeling like an expert is a danger sign, so they deliberately take external courses to leave the internal bubble.
• On resilience, they advise writing your own personal development plan rather than one for your manager, networking by helping without a scorecard, and treating flexibility as the only real job security.
• The post covers Y Combinator's Request for Startups with seven ideas, including Cursor for PMs, AI-native hedge funds, AI-native agencies, stablecoin financial services, AI for government, modern metal mills and AI guidance for physical work.
• It also mentions Google Stitch, a tool that turns a typed idea into a UI and exportable code in seconds.

For brands planning a campaign

Suited to a brief about AI adoption methods. The AI Accelerator Loop is a teachable framework for an internal workshop or co-authored piece on building personal learning loops, while the YC Request for Startups list offers a current map of founder interest for an article on where AI adoption is heading.

They call it "the second job": trying new tools, building small things on your own laptop, and deciding what is real versus noise. They point to Marc Zao-Sanders' 2026 edition of AI in the Wild, a study of 12,637 examples, noting that therapy and companionship ranked above coding and writing, and that almost nobody adopted AI because their employer told them to — a pattern the study names "shadow usage."

The play-understand-share loop: test a tool, work out how it changes your specific workflow, then teach it to your group. A weekly version: pick one tool and give it one hour.

For brands planning a campaign

For AI adoption thought leadership, focus on self-directed experimentation rather than mandatory training. Short, hands-on assets readers can try on their own in an hour fit better than a certification or rollout narrative.

Examples: Source 1 · Source 2

They build on Google Cloud's published list of 1,302 real-world gen AI use cases and draw five trends from it: agents replacing workflows, natural-language querying of legacy systems such as a 40-year-old SAP installation, media production at factory scale, AI reading the physical world, and cybersecurity moving from alerts to action. They extend the argument with an outcomes-as-a-service example from logistics: an agent checks weather, finds suppliers, negotiates, and orders, then simply reports the issue resolved.

Passov is candid about the uncomfortable shift: they felt both lucky and behind at Google Cloud Next, and wrote that pretending the speed is not real helps no one.

For brands planning a campaign

A webinar or article for people seeking advice on AI adoption should demonstrate a workflow that completes without human intervention, not offer prompt-writing tips. Build it around one concrete before-and-after task and the question of who in a company owns the agents doing that work.

Examples: Source 1 · Source 2

They offer three tactical rules: build a personal AI accelerator loop, use writing to turn the day's noise into owned knowledge, and deliberately become a student again before you feel like an expert.

• They also argue for writing your own development plan, helping people without keeping score, and treating flexibility as the only security.
• The line "experience alone doesn't protect you, agility does" recurs in a later post.

For brands planning a campaign

Career and skills messaging for technology professionals can use a reinvention frame rather than a tenure or loyalty story. Readers can act on specific habits: owning your plan, teaching what you learn, and taking an outside course.

Examples: Source 1 · Source 2

Passov relays Udi Ledergor's four-burner stove analogy — health, family, career, social life — with limited gas, so at least one burner has to be turned down; guilt, they write, is the most dangerous background noise for a founder.

• They apply the same logic to themselves, reporting steps up more than 14% through small changes like walking the children to school and taking calls as walk-and-talks.
• They also describe running at 6am in Las Vegas despite jet lag, and leading a living-room Tabata session with their kids on a day when sirens, closed schools, and a cancelled gym pushed everything else aside.

Repeat small fixes rather than making a big plan: protect the asset first, then tackle the task list.

For brands planning a campaign

Brands speaking to founders can pair performance messaging with sustainable-practice framing.

• Passov responds to concrete habits — steps, walks, morning runs — rather than general wellness language.
• Products built around small, repeatable routines fit their approach.

Examples: Source 1 · Source 2 · Source 3

Each issue ends with a single question to reply to and one doable action, such as building a simple agent that runs once without being touched or forwarding the issue to a founder working two jobs. Segments have explicit headers and short numbered lists; the writing moves between a personal anecdote, a named external source, and an instruction.

For brands planning a campaign

Design a sponsored or co-created piece around one segment and one explicit action for the reader. A long brand narrative without a takeaway does not match the structure of the issues in this corpus.

Examples: Source 1 · Source 2 · Source 3 · Source 4

Examples include Y Combinator's Summer 2026 Requests for Startups and their picks from it, Joe Schmidt's a16z article on where it still pays to build, Google Cloud's 1,302 use cases, Anthropic's confidential IPO filing after a raise at a $965 billion valuation, and Mario Gabriele's Saplings, which studied 260 founders across 560 books.

• They also write about organizing the first Founder Private Forum at Google Cloud Next with the AI Futures Fund and Google for Startups, where Jeff Dean joined.
• They take two ideas from the room: the moat is the loop between data, shipped product, and the next training cycle, and founders now plan on two-month cycles.

Pick a small number of items, then add a labeled "my take" instead of summarizing the source neutrally.

For brands planning a campaign

Passov picks up and explains reports, benchmarks, and surveys, giving brands with proprietary data a natural entry point. They link to the original and expect data specific enough to support their own interpretation.

Examples: Source 1 · Source 2 · Source 3

Instead, build memory links for a purchase 12 to 18 months away.

• The three content levels they relay are exclusive insights from proprietary data, original surveys, and polarizing opinions.
• As the post puts it, if you are not making anyone uncomfortable, you are probably not exciting anyone.

Passov credits this framework to their guest rather than presenting it as their own, so the argument comes through a named interview.

For brands planning a campaign

For B2B SaaS audience-building work, Passov can host an interview with a practitioner and translate their advice into specifics.

• Expect credited guest contributions alongside their own commentary.
• To support an exclusive-insights angle, share proprietary data or a survey.

Examples: Source 1

In a conversation with Daniel Mironov, now an investor at Vgames, they report that starting a gaming studio is harder when AI runs through design, code and art: teams are smaller, burn is close to half, and everyone has the same tools, so the edge returns to whether the game holds people.

• Elsewhere, they highlight government, modern metal mills and AI guidance for physical work as the opportunities they find most interesting.
• They argue that the harder go-to-market is itself the moat and point to AI-native service companies that sell an outcome rather than software.

For brands planning a campaign

Passov speaks to vertical AI applications and how operational difficulty can protect a business. A sector campaign should include concrete industry detail: their discussion of these markets draws on named examples and someone else’s firsthand account.

Examples: Source 1 · Source 2 · Source 3 · Source 4

They also write about organizing a founder event at Google Cloud Next '26, consistent with ongoing work there.

View example

Medin describes a software factory where a GitHub issue goes in and a pull request comes out, and a factory where a PRD becomes shipped code.

• Passov describes agents that detect a supply-chain delay, contact compliance, and adjust a forecast, as well as an agent that reorders after checking weather and suppliers.
• Both teach through concrete build examples and first-person experiments rather than abstract predictions.

Medin's material focuses on AI coding agents and the engineering of a reliable autonomous software factory. Passov writes a broader founder-facing newsletter spanning AI adoption, outside research, career strategy, and personal resilience, with less detail on coding specifics.

Read original · Original 2

Brontén asks whether a system guides behavior in the moment or records it afterward, and calls workflows the overlooked lever in sales execution. Passov makes a similar point by contrasting an assistant that drafts an email with an agent that finishes the whole task, and asking who in a company manages the agents doing the work.

Brontén focuses on B2B sales execution, CRM, and workflows around salespeople. Passov covers AI adoption and founder resilience more broadly, touching on sales and marketing through guest interviews.

Read original · Original 2

Soriano gives a re-edit list for turning one UGC clip into five ad tests and a four-number check for diagnosing a stalled ad.

• Passov ends each issue with one question and one task, such as building a simple agent that runs once without being touched.
• Both write in a plain practitioner voice with short numbered steps.

Soriano focuses on paid-social and DTC ad creative, especially first-three-second hooks and creative volume. Passov writes about AI adoption in companies and careers, founder wellbeing, and interpreting outside research for a founder audience.

Read original · Original 2

Their posts are about building an AI software factory: a GitHub issue goes in, a pull request comes out, multiple laps run in parallel, a manager agent checks progress and flags what is stuck, and holdout tests keep the builder from engineering only to visible tests. If the sponsor’s webinar or article needs architecture, coding-agent workflows, reliability practices, or an open-source factory that a technical audience can inspect, Cole’s material is closer to the proof point than Yuval Passov’s broad adoption argument.

The tradeoff is a narrower audience and a different decision.

• Yuval Passov’s posts give operations and product leaders a plain-language shift from passive prompting to managing agents, grounded in published use cases and a founder-facing newsletter format.
• Cole Medin’s posts concentrate on AI coding and agentic software development, with less breadth across AI-adoption research, career and resilience framing, and founder-facing synthesis.
• If the campaign depends on a non-technical adoption narrative, moving to Cole shifts the audience from operations and product leaders toward engineering and product builders seeking implementation detail.

Examples: Source 1 · Source 2

His posts ask what a salesperson will actually work on first in the morning, argue that workflows are the overlooked lever, and test whether a system guides behavior in the moment or just documents it later. A sales-leader audience, or a sponsor selling sales execution software, would get a more direct argument from George than from Yuval Passov.

For a B2B software company’s AI-adoption thought leadership and a webinar or article on moving from passive prompting to managing agents for operations and product leaders, the tradeoff is relevance.

• The campaign decision is about that same need.
• George’s posts stay inside sales performance, CRM, and revenue workflows.
• They do not cover agent management, published AI-use research, or Passov’s founder-and-operator framing.
• Choosing George would change the campaign from “how teams manage agents” to “how sales leaders rebuild seller workflows,” which is a different buyer task even if the AI theme overlaps.

Examples: Source 1 · Source 2

His posts offer a precise diagnostic method for ad creative: compare hook rate, hold rate, CTR, and conversion to find where viewers leave, then re-cut one raw UGC clip into five testable ads by swapping openers, leading with problem or product or proof, changing pace, and rewriting captions. A DTC apparel or e-commerce sponsor that wants a creative testing webinar or article would get a more immediately usable framework from John than from Yuval Passov.

For a campaign on B2B SaaS AI adoption, John is not a substitute.

• His posts address DTC brands, Meta and TikTok creative, and UGC editing, not AI adoption, agent management, operations leadership, or product leaders.
• Switching to John would replace the campaign’s subject and audience.
• He is relevant only if the buyer changes the campaign to performance creative testing.

Examples: Source 1 · Source 2