Giacomo Balli profile picture
Giacomo Balli
Your Consigliere

For owners and founders facing decisions that are expensive to get wrong.
Independent counsel from someone paid only by you.

What’s on Your Mind? LinkedIn

Blog

Writing on technology decisions, AI and buying software: 635 posts since 2012.

274 posts in AI

2026

268 posts
  1. How do Google's AI answers change traffic to a small business website?Google's AI Overviews roughly halve clicks on the searches where they appear, per Pew Research. How to find the queries you lost and write pages that still earn the visit.

    7 min readGrowth

  2. Who is liable when a business's AI agent causes harm?When your AI agent causes harm, your business is usually the one exposed. What the OpenAI and Hugging Face breach shows, what vendor terms say, and five controls to set first.

    7 min readBuying Software

  3. Should my business lock into one AI vendor while prices keep falling?Anthropic and OpenAI cut frontier AI prices by 20% to 50% in one week. Why owners should keep AI contracts short, stay model-agnostic, and re-test every year.

    7 min readBuying Software

  4. What Happens When Users Stop Looking for Apps?For the last 15 years, software discovery has followed a fairly familiar pattern. You realize you need something, you search Google , maybe check Redd

    5 min readAI

  5. OpenAI is commoditizing successful AI startups unusually fastOpenAI is commoditizing successful AI startups unusually fast: Meta Muse XAI Grok Bot → Dots.TypeSafe AI Jev → Decisions API.Agent frameworks → Codex...

    1 min readAI

  6. AI-built product security worries usually land on the wrong layerWhen an owner worries about the security of an AI-built product, the worry usually lands on the wrong layer.

    1 min readAI

  7. Your most exposed IP may be how your company knows what to doYour most exposed IP may be how your company knows what to do. Companies worry about sending the confidential PDF, the source code or the customer list to...

    2 min readAI

  8. The same question comes round the table every few monthsThe same question comes round the table every few months wearing different clothes. Should we build it. Should we replace it.

    1 min readAI

  9. Companies may be about to fire exactly the wrong peopleCompanies may be about to fire exactly the wrong people. Not because AI can’t do a lot of their work.

    2 min readAI

  10. You cannot model a software build on the vendor's estimateA fractional CFO asked me how to model a software build in the plan when the only number available is the vendor's estimate.

    1 min readBuying Software

  11. Before hiring an AI consultancy, find the part that actually needs AIBefore hiring an AI consultancy, work out which part of the problem actually requires AI. It is usually smaller than the proposal.

    1 min readAI

  12. When is an AI API wrapper enough, and when do you need more?An API wrapper around ChatGPT or Claude is enough for generic, reviewed, low-risk work. You need more once the AI speaks for your company, uses your data, or is the product.

    8 min readBuying Software

  13. A $5,000 technology decision can deserve more scrutiny than a $500,000 oneA $5,000 technology decision can deserve more scrutiny than a $500,000 one. Owners scrutinise by price, and price is a poor guide to consequence.

    1 min readAI

  14. A CEO sent me a technology proposalA CEO sent me a technology proposal. These are the first three things I ignored. The feature list.

    1 min readBuying Software

  15. When should a CEO get a second opinion on a technology decisionNot on all of them. There are thresholds, and they are worth writing down before the next proposal arrives, because after it arrives the answer is always...

    2 min readBuying Software

  16. The person recommending the technology should not be the only one valuing itThe person recommending the technology should not be the only person estimating its value.

    2 min readAI

  17. The most expensive sentence in a software project: "We'll figure that out later"The most expensive sentence in a software project is "we'll figure that out later." It is also, sometimes, the right call.

    1 min readAI

  18. The best model is becoming a surprisingly bad defaultThe best model is becoming a surprisingly bad default. I was about to tokenmaxx my Claude weekly quota, so I switched to Sonnet 4.5.I expected a noticeable...

    1 min readAI

  19. AI made software dramatically cheaper to createAI made software dramatically cheaper to create. It did not make bad software decisions cheaper. If anything, it made them more common.

    1 min readAI

  20. The prototype took three weeksThe prototype took three weeks. The production system is quoted at nine months. Owners read those two numbers and conclude someone is lying.

    2 min readAI

  21. The $180,000 dispatch system and the question about month five"The $180,000 dispatch system will take four months. Before AI it would have been nine." Fine. Ask what happens in month five.

    1 min readAI

  22. The $20 AI tool versus the $200,000 AI implementationThe $20 a month AI tool and the $200,000 AI implementation are often pitched for the same problem, and the honest answer about which one you need is usually...

    1 min readAI

  23. Your team just proposed an AI initiativeYour team just proposed an AI initiative. These are the questions I would ask before approving the budget, in the order I would ask them.

    1 min readAI

  24. The dangerous AI project is the one that worksThe dangerous AI project is the one that works. Failures are cheap. The working one stays.

    1 min readAI

  25. "We need an AI strategy" is usually the wrong first sentence"We need an AI strategy" is usually the wrong first sentence, and it produces a predictable document: a list of AI capabilities with your company's name...

    1 min readAI

  26. The questions a CEO should ask in the approval meetingThe CEO does not need to understand the technology. They need to know which questions expose a bad decision.

    1 min readAI

  27. Three proposals for the same project: $80k, $240k and $600kThree proposals for the same project: $80k, $240k and $600k. They are for the same project the way three quotes to "fix the roof" are for the same roof.

    2 min readBuying Software

  28. Share of new apps with "AI" in the name, by year: 0.1% through 2016Share of new apps with "AI" in the name, by year: 0.1% through 2016. 6% so far in 2026. The peak was last year.

    1 min readAI

  29. The expertise that built an institution can quietly become the reason it stopsThe expertise that built an institution can quietly become the reason it stops learning. Past success gives leaders judgment and pattern recognition, but it...

    1 min readAI

  30. Years of internal operational mess are about to have a market valueCompanies are about to discover that years of internal operational mess have a market value.

    1 min readAI

  31. AI changes the backlogAI changes the backlog. Economically:Lower implementation cost → lower experimentation threshold → more experiments → more discovered demand.

    1 min readAI

  32. The Startup TrapA startup has ten employees and six months of runway. A large prospective customer appears and offers a contract worth $500,000, but wants a few chang

    10 min readAI

  33. Beautiful software architecture historically mattered because humans neededBeautiful software architecture historically mattered because humans needed to understand and modify systems cheaply.

    1 min readAI

  34. 37signals let designers use agents to implement features37signals let designers use agents to implement features. Individual PRs looked reasonable, but collectively they degraded the architecture enough that...

    1 min readAI

  35. The best AI for your business probably won’t be the smartest oneThe best AI for your business probably won’t be the smartest one. The big, powerful AI models will keep pushing the limits.

    1 min readAI

  36. NVIDIA buying Hugging Face could make open AI depend on a single vendorNVIDIA buying Hugging Face could make “open AI” more dependent on a single vendor.Reports put the potential deal around $13 billion.

    1 min readBuying Software

  37. Excellence pays only where it's visibleExcellence pays only where it's visible. collected what 32 people who build at the frontier of AI think work looks like after automation.

    2 min readAI

  38. Nobody hacked Hugging FaceNobody hacked Hugging Face. At Black Hat this month, OpenAI presented what happened when they turned unconstrained agents loose on a cybersecurity evaluation.

    1 min readAI

  39. I brought in Codex as an adversarial critic of my thinking with ClaudeI was worried I was delegating too much of my thinking to Claude so I brought in Codex as an adversarial critic.

    1 min readAI

  40. AI demos work because the data is cleanAI demos work because the data is clean. Your data isn't, and that's the actual reason the pilot underwhelmed.

    1 min readAI

  41. Your company may own a valuable data asset that has never appeared on yourYour company may own a valuable data asset that has never appeared on your balance sheet. Every quote you sent, customer you lost, support issue you solved...

    1 min readAI

  42. Everyone's talking about agentsEveryone's talking about agents. Almost nobody has written down what theirs is allowed to do. An agent isn't a smarter chatbot.

    1 min readAI

  43. "We don't have anyone who knows how to do this." Neither did the companies"We don't have anyone who knows how to do this." Neither did the companies that figured it out.

    1 min readAI

  44. Most AI pilots produce nothing measurableMost AI pilots produce nothing measurable. That's not a technology problem. It's a scoping problem, and it looks like this: the pilot starts with "let's see...

    1 min readAI

  45. "What if I get a $4,000 bill?" This is the most solvable fear on the list"What if I get a $4,000 bill?" This is the most solvable fear on the list and almost nobody addresses it before signing up.

    1 min readAI

  46. "I don't trust it to be right." You shouldn't"I don't trust it to be right." You shouldn't. That's the correct instinct and it's the wrong conclusion. The mistake is framing this as trust vs no trust.

    1 min readAI

  47. "I can't put customer data into an AI tool." Correct"I can't put customer data into an AI tool." Correct. Most owners hear "AI" and picture their entire CRM getting uploaded somewhere.

    1 min readAI

  48. A $150k employee no longer costs $150kA $150k employee no longer costs $150k. AI is making employee cost dynamic. Salary is only the starting point.

    2 min readAI

  49. Everyone seems focused on when we’ll reach AGIEveryone seems focused on when we’ll reach AGI.But I think we’re missing the more immediate point: the AI we already have is capable of doing a huge portion...

    1 min readAI

  50. Newer AI models cost more per token but often still cheaper to useNewer AI models cost more per token but often still cheaper to use. Because smarter models often need fewer tokens to get the job done.

    1 min readAI

  51. Building software is no longer the hard part of starting a companyBuilding software is no longer the hard part of starting a company. That's exactly why more founders are going to build products nobody wants.

    1 min readAI

  52. Signs your business has outgrown its off-the-shelf softwareThe clearest signs you have outgrown a generic tool are spreadsheets duct-taping gaps, unused seats, manual workarounds, broken integrations, and data trapped in silos.

    6 min readBuying Software

  53. How to make AI "know everything" about your businessI often get asked by companies how they can make AI "know everything" about their business.

    2 min readAI

  54. The price of content is about to separate from the price of accessThe price of content is about to separate from the price of access.For years, publishers monetized distribution.

    1 min readAI

  55. The website stopped being the scarce assetThe website stopped being the scarce asset. Merchant identity is. As more purchases begin inside AI assistants instead of storefronts, the question shifts...

    1 min readAI

  56. Owning your own e-commerce store has become cheaper than earning attentionOwning your own e-commerce store has become cheaper than earning attention inside someone else’s marketplace.

    1 min readAI

  57. Entry-level jobs are becoming a less reliable way to predict future headcountEntry-level jobs are becoming a less reliable way to predict future headcount. The common assumption is simple: if AI can do part of the work, companies...

    2 min readAI

  58. People are confusing "can build" with "should own"A lot of people are confusing “can build” with “should own.” The argument goes like this: AI makes software development dramatically cheaper, therefore...

    2 min readAI

  59. Everyone talks about how AI makes startups easier to buildEveryone talks about how AI makes startups easier to build. That’s true, but it’s not the most important change.

    1 min readAI

  60. X (formerly Twitter) started asking Claude for moneyX (formerly Twitter) started asking Claude for money. Discuss on LinkedIn

    1 min readAI

  61. I spend a lot of time helping business owners think through AI strategyI spend a lot of time helping business owners think through AI strategy. One thing that keeps coming up is that we've been assuming AI agents will use...

    2 min readAI

  62. Everyone seems to be reading this chart from Financial TimesEveryone seems to be reading this chart from Financial Times as: "AI is creating lots of apps nobody wants." That's not what the research says.

    1 min readAI

  63. At #WWDC26, Apple finally delivered much of what it had promised two yearsAt #WWDC26, Apple finally delivered much of what it had promised two years earlier. Many people will look at that and conclude Apple was behind on #AI.

    2 min readAI

  64. The people most obsessed with startup ideas rarely start a companyThe people most obsessed with finding startup ideas are often the least likely to start a company.

    2 min readAI

  65. The next defensible advantage may come from process data, not customer dataThe next defensible advantage may come from process data, not customer data. Customer databases are becoming easier to replicate than most executives want...

    1 min readAI

  66. UI is the worst thing technologists builtUI is the worst thing technologists built. Software companies organize themselves around screens.

    1 min readAI

  67. Analytical SaaS is losing the right to charge for answersAnalytical SaaS companies are losing the right to charge for answers that can be generated directly from the underlying data.

    1 min readAI

  68. 120 employees, 40 experimenting with AI, all trapped in private tabs120 employees.40 experimenting.And it’s all trapped in private tabs. This isn’t an AI problem. It’s a knowledge management problem.

    1 min readAI

  69. Most firms are optimizing the wrong variableMost firms are optimizing the wrong variable. They’re treating AI like a cheaper labor substitute: reduce cost, automate steps, accelerate workflows.

    1 min readAI

  70. AI is unbundling professional services by deleting the middle of the workAI is unbundling professional services by deleting the middle of the work. The “judgment sandwich” is becoming obvious.

    1 min readAI

  71. “10x engineering” is a comforting story because it gives you a single lever“10x engineering” is a comforting story because it gives you a single lever: ship more code.

    2 min readAI

  72. In some companies you’ll see headcount traded for tokensIn some companies you’ll see headcount traded for tokens. Not because leaders suddenly prefer “AI” to humans, but because tokens look like a variable cost...

    2 min readAI

  73. I grew up in ItalyI grew up in Italy. I live in the US now. Same human problem, two opposite answers, and I've watched both up close.

    2 min readAI

  74. The most effective way to disrupt an incumbent is to remove the problemThe most effective way to disrupt an incumbent is to remove the problem it was created to solve. OpenAI Sites feels like another step in that direction.

    1 min readAI

  75. Domain expertise has always been the real moat because it is the only advantageDomain expertise has always been the real moat because it is the only advantage you cannot buy on a roadmap.

    1 min readAI

  76. When AI Makes Features Cheap, What Becomes the Moat?A decade ago, a small mobile team could spend months building something competitors could not easily replicate. Today a capable developer with modern

    9 min readAI

  77. Most people still think of Shopify as a website builderMost people still think of Shopify as a website builder. That version of Shopify barely exists anymore.

    1 min readAI

  78. I’ve been in mobile since 2010I’ve been in mobile since 2010. ASO has always been part of the game, so over the years I’ve tried nearly every tool that entered the market.

    1 min readAI

  79. AI adoption fails when learning stays trapped inside individual workflowsAI adoption fails when learning stays trapped inside individual workflows. A developer at Siemens may find a better way to use agents in code review.

    1 min readAI

  80. Agentic code is free like a baby goatAgentic code is free like a baby goat.The first day is the cheapest day you will ever have with it.

    1 min readAI

  81. The next platform shift will not be won where users clickThe next platform shift will not be won where users click, it will be won where agents connect.

    1 min readAI

  82. Freezing junior engineering hiring is not a cost moveFreezing junior engineering hiring is not a cost move, it is an organizational risk decision disguised as efficiency.

    1 min readAI

  83. AI cost conversations usually start at the model and end at pricingAI cost conversations usually start at the model and end at pricing. That is the least relevant part of the system.

    2 min readAI

  84. For decades we've shipped frozen interfaces where every pixel was decidedFor decades we've shipped frozen interfaces where every pixel was decided by a designer years before you opened the app.

    1 min readAI

  85. AI is being discussed as if the economy is mostly software, contentAI is being discussed as if the economy is mostly software, content, and knowledge work. That framing holds if your world is screens, documents, and APIs.

    2 min readAI

  86. One careless download can cascade across companiesOne careless download can cascade across companies. In February 2026, an employee at Context.ai installed a Roblox cheat.Hidden inside: Lumma Stealer malware.

    1 min readAI

  87. Moats are retention tools sold as growth strategyMoats are retention tools sold as growth strategy. Switching costs and data gravity make leaving expensive.

    1 min readAI

  88. Anthropic has figured out that fear sellsAnthropic has figured out that fear sells. They frame their releases in existential terms, leaning into “this could be dangerous” narratives.

    1 min readAI

  89. In AI work, volume is not vanityIn AI work, volume is not vanity. It is a control surface. Most organizations treat quantity as a cost center.

    1 min readAI

  90. Engineering teams that went heaviest on AI coding tools are now the onesEngineering teams that went heaviest on AI coding tools are now the ones struggling most with stability.

    1 min readAI

  91. Most “disruption” is just a better interface on top of the same systemMost “disruption” is just a better interface on top of the same system. Real disruption is when the incumbent’s operating model becomes incompatible with...

    1 min readAI

  92. People often ask if you still need to be in San Francisco to build somethingPeople often ask if you still need to be in San Francisco to build something meaningful in tech. For a while, it looked like the answer might be no.

    2 min readAI

  93. I built a Python script that generates App Store screenshots in 1sI built a Python script that generates App Store screenshots in 1s. Drop in a screenshot, get back a polished image with device bezel, 3D tilt, text...

    1 min readAI

  94. If your ARPU did not go up by at least 50%If your ARPU did not go up by at least 50%, your new AI component is just a feature. Inside most companies, AI lands where the org can absorb it: a workflow...

    1 min readAI

  95. AI boosted coding and now is moving to GTMAI boosted coding and now is moving to GTM. GTM cannot fix PMF. It can only scale whatever truth the product already has.

    1 min readAI

  96. Apple will finally have their own niche hyper-performant secure modelsMost companies think the AI race is about building the biggest model. In practice it is about who controls the smallest ones.

    1 min readAI

  97. The highest paid AI roles will belong to the people who can makeThe highest paid AI roles will belong to the people who can make the organization change its behavior, not the people who can make the model smarter.

    1 min readAI

  98. Life with ClaudeLife with Claude. Discuss on LinkedIn

    1 min readAI

  99. The AI job everyone is talking about does not existThe AI job everyone is talking about does not exist. “Prompt engineer” was a temporary phase.

    2 min readAI

  100. I keep seeing the same mistake from vibecodersI keep seeing the same mistake from vibecoders: they describe iOS components instead of naming them.

    1 min readAI

  101. Everyone is talking about “AI agents replacing ecommerce.”Everyone is talking about “AI agents replacing ecommerce.” Reality check: the first real experiment didn’t go well.

    1 min readAI

  102. White House Unveils National AI Policy FrameworkWhite House Unveils National AI Policy Framework Speed first, regulation later. Discuss on LinkedIn

    1 min readAI

  103. One of the stranger patterns in AI adoption is emotional, not technicalOne of the stranger patterns in AI adoption is emotional, not technical. Lower and middle-income regions often read AI as upside.Wealthier regions read it...

    1 min readAI

  104. When a company tells me “AI is helping people grieve, learn, get careWhen a company tells me “AI is helping people grieve, learn, get care, access services”, I believe them.

    1 min readAI

  105. Anthropic just changed how AI coding tools will be usedAnthropic just changed how AI coding tools will be used. They introduced Claude Code channels.

    1 min readAI

  106. Most AI risk conversations are framed like this:Most AI risk conversations are framed like this: “There is the value, and then there are the harms.” That framing is comforting.

    2 min readAI

  107. Most leaders are mispricing AIMost leaders are mispricing AI. They’re arguing about “intelligence” while the defensible value is a service bundle humans can’t deliver at scale: patience...

    1 min readAI

  108. Most AI conversations inside companies start with: “We need to be faster.”Most AI conversations inside companies start with: “We need to be faster.” Anthropic ran 81,000 interviews and kept hearing the same pattern: people begin...

    1 min readAI

  109. Stripe is launching the Machine Payments Protocol (MPP) open standardStripe is launching the Machine Payments Protocol (MPP) open standard. Most leaders still treat payments as a “checkout problem”.

    1 min readAI

  110. Racing used to punish mediocrityRacing used to punish mediocrity. With a combustion engine, the car was a system you had to actively manage:gear selection, power band, traction, cooling...

    2 min readAI

  111. Most operators say they want teams that “understand the system.”Most operators say they want teams that “understand the system.” What they consistently reward is speed of response inside it.

    1 min readAI

  112. Serial founders talk about "batting average" like the goal is to never strikeSerial founders talk about "batting average" like the goal is to never strike out. That is a misunderstanding of the game they are actually playing.

    1 min readAI

  113. "Agents" without an orchestrator are just new tools with a chat UI."Agents" without an orchestrator are just new tools with a chat UI. In most non-tech companies, the hard part is not getting a model to take an action.

    1 min readAI

  114. AI does move fast.AI does move fast. But the thing that leaves senior managers behind is not the pace of models. It is the pace of decision loops.

    1 min readAI

  115. There is a quiet grief a lot of builders are carrying right now.There is a quiet grief a lot of builders are carrying right now. Not fear of AI taking jobs. The loss of a specific kind of joy.

    1 min readAI

  116. Was listening to Carlos Gonzalez de Villaumbrosia (Product School) and TimWas listening to Carlos Gonzalez de Villaumbrosia (Product School) and Tim Simmons (Walmart) and something really caught my attention: Before AI Agents and...

    1 min readAI

  117. Everyone tells the ATM story wrong.Everyone tells the ATM story wrong. They treat labor substitution like a task list: automate the cash withdrawal, the teller disappears.

    1 min readAI

  118. With AI agents, complexity stops being a tax and starts becoming a moatWith AI agents, complexity stops being a tax and starts becoming a moat. Most leaders still treat complexity like something to "simplify away": fewer...

    2 min readAI

  119. AI is going to create job categories that sound fake until you need one tomorrowAI is going to create job categories that sound fake until you need one tomorrow. Not because the tools get better.

    2 min readAI

  120. The early labor signal from AI is not layoffs.The early labor signal from AI is not layoffs. It is the hiring freeze you do not call a freeze.

    2 min readAI

  121. Most SaaS is designed to be chosen by a person who is slightly unsure.Most SaaS is designed to be chosen by a person who is slightly unsure. So we obsess over UX polish, onboarding flows, "time to value", and the story you can...

    2 min readAI

  122. AI Is Commoditizing Intelligence.AI Is Commoditizing Intelligence. Value Is Moving Elsewhere. AI capability is becoming widely accessible.

    2 min readAI

  123. Most organizations are thinking about AI backwards.Most organizations are thinking about AI backwards. They start with, "Where can we add AI?" and that almost always produces the same roadmap: assistants...

    2 min readAI

  124. AI isn't making "development" more valuable.AI isn't making "development" more valuable. It's making raw output less scarce. When a studio can ship 3x the code with the same headcount, the client's...

    1 min readAI

  125. The #SaaSpocalypse is a great narrative because it takes a real shift and turnsThe #SaaSpocalypse is a great narrative because it takes a real shift and turns it into a clean conclusion.

    1 min readAI

  126. Really Apple?This was such low hanging fruit... I tried the new "Intelligence" in Xcode, hoping it would have a deeper integration than Claude Code (Anthropic) or Codex...

    1 min readAI

  127. Most clients don't ask for clarity.Most clients don't ask for clarity. They ask for solutions. "We need a mobile app." Maybe an app is the right answer.

    1 min readAI

  128. The theory that we'll vibe code everything is 'flat wrong'.The theory that we'll vibe code everything is 'flat wrong'. --Anish Acharya If you think AI is going to replace your engineers, you are about to create the...

    1 min readAI

  129. "Digital transformation" is not disruption."Digital transformation" is not disruption. Buying software does not change your market position. It just changes what you pay for, and who you depend on.

    1 min readAI

  130. AI didn't just speed up my work.AI didn't just speed up my work. It changed where the work lives. At first, AI helped me do the thing. Faster execution.

    1 min readAI

  131. Coding with AI doesn't reduce rigorCoding with AI doesn't reduce rigor. It increases the penalty for skipping it. The more I use AI for code, the clearer this gets: Tests stop being optional.

    1 min readAI

  132. AI isn't replacing workers.AI isn't replacing workers. It's replacing what work even means. Most people still frame AI as productivity software, or task-completing agents.

    1 min readAI

  133. Sometimes the reason the category is crowded is the same reason it will notSometimes the reason the category is crowded is the same reason it will not support a durable business.

    1 min readAI

  134. AI will replace the parts of consulting that look like consulting.AI will replace the parts of consulting that look like consulting. So why do companies still pay for the rest?

    1 min readAI

  135. Your data moat gets weaker the moment interpretation gets cheap.Your data moat gets weaker the moment interpretation gets cheap. So what happens when your "advantage" was just owning the spreadsheet?

    2 min readAI

  136. The biggest impact of LLMs is not replacing software.The biggest impact of LLMs is not replacing software. It is turning "unwritten work" into something a business can actually run.

    1 min readAI

  137. AI will commoditize the part of software your CFO obsesses over.AI will commoditize the part of software your CFO obsesses over. So why are teams still choosing vendors based on features and day rates?

    1 min readAI

  138. The internet did not kill publishing. It broke the unit of value.The internet did not kill publishing. It broke the unit of value. If code generation gets "free", what exactly will SaaS customers still be paying for?

    1 min readAI

  139. AI makes writing and analysis cheap.AI makes writing and analysis cheap. So what are you still paying humans for? Most operators respond by layering tools onto the current process.

    1 min readAI

  140. Tokens are not the bottleneck. Humans are.Tokens are not the bottleneck. What happens to "AI adoption" when nobody has to ask? Most companies treat AI like a new UI layer.

    1 min readAI

  141. AI did not kill software engineering. It moved the mistakes upstream.AI did not kill software engineering. It moved the mistakes upstream. If code is cheap now, why do so many teams still ship the wrong thing faster?

    1 min readAI

  142. Law is just the first industry to break (thanks to Harvey and Legora).Law is just the first industry to break (thanks to Harvey and Legora). Every knowledge job is next.

    2 min readAI

  143. Most SaaS companies won't die because of competitionMost SaaS companies won't die because of competition. They'll die because their pricing model stops making sense.

    2 min readAI

  144. AI is not blocked by intelligence. It's blocked by org charts.AI is not blocked by intelligence. It's blocked by org charts. Everyone thinks AI adoption is about better models.

    1 min readAI

  145. A strong architect is paid to say "no" earlier than everyone else.A strong architect is paid to say "no" earlier than everyone else. When was the last time your "senior" decision reduced options instead of adding features?

    1 min readAI

  146. One old Cordova app (first launched on Aug 4, 2011).One old Cordova app (first launched on Aug 4, 2011). Three days and now fully migrated to native Swift for iOS.

    1 min readAI

  147. AI makes products look simpler while making them harder to run.AI makes products look simpler while making them harder to run. When a UI collapses into a prompt box, leaders assume complexity disappeared.

    1 min readAI

  148. PMF is no longer something you "achieve"PMF is no longer something you "achieve"; it's something you re-earn continuously. If last month's PMF can expire, what exactly were you congratulating...

    1 min readAI

  149. Data advantage is no longer about owning data.Data advantage is no longer about owning data. If data is truly the moat, why does everyone also expect it to be portable?

    1 min readAI

  150. The real scarcity is judgment, not execution.The real scarcity is judgment, not execution. If building is cheap now, why do so many digital programs get slower?

    1 min readAI

  151. Customer Support: Growth Lever or Labor Sponge?Most operators hear the Zappos Family of Companies story and take the safe lesson: invest in people, create loyalty, win on experience.

    2 min readAI

  152. How Companies Will Hire Based on AI UsageSmart companies will hire you based on "how" you use AI. McKinsey & Company started piloting an interview where candidates use its internal AI tool, Lilli...

    1 min readAI

  153. Understanding VC Investment Stages: From Seed to DominanceTo all the VC people... does this sound correct? Seed: “Does this deserve to exist?” You are underwriting judgment under uncertainty and early signals that...

    1 min readAI

  154. How a 10% Credit Card APR Cap Could Raise Your Growth CostsA 10% cap on credit card APR can raise your growth costs. If the rate goes down, why would acquisition get harder?

    1 min readAI

  155. AI in Engineering, Leadership, and SalesClarity is not something you find before you move. It is something you earn by moving. So why do we keep waiting for a cleaner vision before we act?

    1 min readAI

  156. GEO: The New Frontier Beyond SEO and AEOGEO is not a marketing tactic. It is an operations problem. Who owns the truth when the customer never clicks?

    2 min readAI

  157. Why Education No Longer Guarantees EmploymentEducational institutions are no longer a reliable path to employment. It is a bet that companies may never need to honor.

    2 min readAI

  158. Drone Delivery Is Easy. Getting Picked by AI Is Hard.Drone delivery is the easy part. Getting picked by the AI that starts the shopping journey is harder.

    1 min readAI

  159. Why Agentic Checkout Will Break Your Business RulesAgentic checkout will not break because the AI is wrong. It will break because your business rules are.

    1 min readAI

  160. Stop Wasting $100K on Dev Shops Without StrategyWant to throw away $100k? just tell a dev shop to build your idea. 😂 Most founders rush to hire a developer before they have a strategy.

    2 min readAI

  161. The Ugliest Process Is Usually Keeping You AliveThe ugliest process in a company is usually the one keeping it alive. Before you delete it, do you know what risk it was built to absorb?

    1 min readAI

  162. AI Output Quality Mirrors Your Operating DisciplineThe quality of your AI output is a mirror of your operating discipline. Are you using a model, or are you outsourcing requirements?

    1 min readAI

  163. Why You Should Use Plan Mode Before Writing AI CodeThe fastest way to get bad AI code is to start typing immediately. Have you decided what “done” looks like, or are you outsourcing the thinking too?

    1 min readAI

  164. Most Pricing Problems Are Discovery ProblemsMost pricing problems are really discovery problems. “Worth” is not something you carry into the room.

    1 min readAI

  165. AI in Customer Support, Ecommerce, and LeadershipFor years, my default advice was “go hybrid first”. When a founder had 50k, a deadline, and no mobile team, Cordova was a rational way to ship something...

    1 min readAI

  166. Most product ideas fail at the feasibility layer, not the idea layerMost product ideas fail at the feasibility layer, not the idea layer. Time, budget, team, and sequencing quietly invalidate “great” ideas all the time.

    1 min readAI

  167. Why Forward-Deployed Engineers Are the New Operating SystemForward-deployed engineers are now the operating system. Palantir Technologies won early because it put people in the messy middle, translating “the model...

    1 min readAI

  168. Your AI Strategy Will Break With the Next Model ReleaseYour AI “strategy” will look brilliant right up until the next model release breaks it. I keep seeing operators in logistics, insurance, and industrial...

    1 min readAI

  169. Why Remote Work Fails for Sense-MakingThe project is “on track” for six weeks, then suddenly you cannot explain what you are building to a customer.

    1 min readAI

  170. Why Early-Stage Founders Don't Need a Full-Time CTOMost founders do not fail because they cannot build. They fail because they buy the wrong build.

    2 min readAI

  171. Why Most GenAI Rollouts FailMost GenAI rollouts fail for the same reason the first ML projects failed. Everyone budgeted for a subscription and got a systems integration.

    1 min readAI

  172. Enterprise and Data with AIWe used to ship POS and loyalty updates in 7-10 days. Then we added a “lightweight” process and it became 6 weeks.

    1 min readAI

  173. AI in Product, Delivery, and DataWe used to blame our “slow” team for missing the loyalty app launch. Then we added two more squads and missed by another quarter.

    1 min readAI

  174. AI, Customer, and HiringI never watched a founder hand over their gut to a dashboard. Then the dashboard started running the company.

    1 min readAI

  175. AI in Customer Support, Product, and LeadershipI used to applaud teams for “taking risks.” Then I watched the same launch fail three times with three different names.

    1 min readAI

  176. AI and Customer SupportI used to think senior engineers earned trust by always having an answer. Then I watched a launch slip because nobody wanted to say “I do not know.” At a...

    1 min readAI

  177. AI in Engineering, Pricing, and DataWe did not rewrite anything. Checkout still dropped from 9 seconds to 5. A regional grocery chain had a loyalty app and a small ops team.

    1 min readAI

  178. AI, Product, and DeliveryWe never tracked “delivery speed” until the CEO asked for one number. Two sprints later, we were “faster” and member complaints doubled.

    1 min readAI

  179. AI in Automation, Engineering, and EducationPerhaps the real AI investment for businesses is deciding what NOT to automate. Your AI pilot can “succeed” and still stall the business.

    1 min readAI

  180. Strategy and Enterprise with AIAI is replaying the SaaS cycle in reverse. “ChatGPT for everything” is the easy part. A mid-market insurer rolled out a generic AI chat tool to help...

    1 min readAI

  181. mobile apps do notMost mobile apps do not fail in the big demo. They fail at 8:12am, on a spotty network, when the store opens.

    1 min readAI

  182. AI in Engineering, Hiring, and DataI went home to Italy for Christmas. On my flight back to SF, I watched Jersey Boys and it reminded me that talent does not protect you from a bad deal.

    1 min readAI

  183. I have an old Cordova app for iOS I wanted to convert to nativeI have an old Cordova app for iOS I wanted to convert to native (Swift for Apple iOS). I tasked both Codex (OpenAI) and Claude (Anthropic) to generate a...

    1 min readAI

  184. This year reinforced something I have seen over and over across productsThis year reinforced something I have seen over and over across products, companies, and founders. Effort is rarely the constraint.

    1 min readAI

  185. AI in Automation, Ecommerce, and DeliveryMost grocery apps fail for a boring reason. Nobody agreed on the one number it should move. Margins are thin and shoppers bounce fast.

    1 min readAI

  186. AI code assistants do not make your team fasterAI code assistants do not make your team faster. They make your weak spots louder. In property and casualty insurance, the moment you point GitHub Copilot...

    1 min readAI

  187. The 10x productivity claim with AI Is still misThe 10x productivity claim with AI Is still misleading but in the way you're thinking. Saying AI makes your devs "10x engineers" hides the real variable...

    2 min readAI

  188. If I see one more person calling Generative AI a toolIf I see one more person calling Generative AI a tool, my brain is gonna start seeping out of my ears.

    2 min readAI

  189. Customer Support and Strategy with AII asked AI to tell me about how I use AI: Here's what it had to say: "You use AI primarily as a leverage tool for thinking, not as a substitute for it...

    1 min readAI

2025

94 posts
  1. Customer Support and Enterprise with AII wonder how much of what we call “finance best practice” is really just coping with banks being closed. Your month-end close is a map of bank cutoffs.

    1 min readAI

  2. One of my first apps is finally changing handsOne of my first apps is finally changing hands. I released it in 2010. Fifteen years ago. Long before subscriptions existed, before TestFlight existed...

    1 min readAI

  3. Most “privacy-first” products still know your real identityMost “privacy-first” products still know your real identity. Email reset, IP logs, phone checks - that is a dossier.

    1 min readAI

  4. AI, Customer, and PricingYour next SaaS renewal should feel harder than last year. That is not procurement being annoying. It is the math changing.

    1 min readAI

  5. AI in Product, Engineering, and DataApple will reject your iPhone app for one shady permission. Google Chrome will “Feature” a browser extension that watches your AI chats.

    1 min readAI

  6. Most people get competition wrongMost people get competition wrong. They look at features and direct substitutes. Cal AI vs MyFitnessPal vs Lose It!

    1 min readAI

  7. AI and StrategyMost big consulting engagements are not bought for insight. They are bought for liability transfer.

    1 min readAI

  8. Operations and Data with AIMost grocery stores do not need a data center to decide if a shelf is empty. That is why NPUs are showing up in checkout lanes, handhelds, cameras, and...

    1 min readAI

  9. Growth and Pricing with AIMost big private companies avoid the public markets because they can (Stripe, ByteDance, SHEIN, Anthropic, Canva, Fanatics etc).

    1 min readAI

  10. AI, Customer, and EnterpriseMost bank AI bans are really policy statements, not behavior changes. The work just goes underground.

    1 min readAI

  11. AI, Customer, and DeliveryDomain knowledge is the only moat. AI can write the code, but it cannot own the mess it creates.

    1 min readAI

  12. AI, Growth, and DeliveryThe Netflix lawsuit gets one thing wrong right from the start: it’s fighting a market that no longer exists.

    2 min readAI

  13. Rethinking AI (8)Everyone is terrified of being wrong. That’s a shame, because being wrong is one of the most useful signals you can get.

    1 min readAI

  14. Customer Support and Automation with AIRecent discussions around AI in radiology often suggest that automation will replace human expertise.

    1 min readAI

  15. AI, Engineering, and StartupsWe talk about AI as if it’s a contest of smarter models. What’s actually unfolding is a clash of long-standing institutional playbooks...

    2 min readAI

  16. The loud debates about AI safety tend to focus on sci-fi risThe loud debates about AI safety tend to focus on sci-fi risks. The quieter, more immediate risk looks very different: a world where one company ends up...

    1 min readAI

  17. corporate transformationsMost corporate transformations fail for a simple reason: they’re a substitute for thinking, not a product of it.

    2 min readAI

  18. AI in Operations, Growth, and EngineeringEveryone keeps saying content is king. – Paramount lacks scale. – The Walt Disney Company struggles with identity.

    2 min readAI

  19. AI and GrowthEveryone is arguing about whether Google's AI models are “better” but that misses the real advantage entirely.

    1 min readAI

  20. Fast cheap good. Pick two?Fast cheap good. Pick two? That line should have retired with flip phones. I help founders shape the strategy, the roadmap, and the calls that decide...

    1 min readAI

  21. AI in Engineering, Delivery, and DataAI assisted coding isn't a "boost" anymore. It is the default way devs ship software (or your team falls behind).

    1 min readAI

  22. AI in Growth, Engineering, and LeadershipMost CEOs are optimizing for a world that no longer exists. Progress isn’t driven by resources, talent, or even technology.

    2 min readAI

  23. AI and Hiring"AI Can Technically Perform 12% of US Labor Market's Wage Value" via Massachusetts Institute of Technology Most leaders point AI at dev teams.

    1 min readAI

  24. The bar for “good enough” isThe bar for “good enough” is dead. The new bar is delight. AI has democratized building apps.

    1 min readAI

  25. Big tech just missed the real story: satellitesBig tech just missed the real story: satellites are about to reshape mobile. Most people still treat Starlink (SpaceX) as rural-internet plays.

    2 min readAI

  26. AI and EnterpriseMost SaaS companies are protecting the wrong moat. We’ve spent years thinking features, UI polish, and enterprise checkboxes keep competitors at bay.

    1 min readAI

  27. AI in Product, Sales, and DataAI advertising is about to hit an inflection point, and almost nobody is talking about the real reason why.

    2 min readAI

  28. AI in Product, DataLet's stop with the self-help guilt. That feeling that you can't focus on anything long enough to get real work done isn't a discipline problem.

    1 min readAI

  29. AI in Automation, Product, and GrowthCoding just ate its own tail... and it tastes like rocket fuel. Imagine building the tool that makes software engineers 10x faster… while your entire...

    2 min readAI

  30. Automation and Growth with AIMost AI tools today are sold as “cost savers”: shave 20% off headcount, automate one step in a five-step process, done.

    1 min readAI

  31. AI in Automation, Strategy, and DataThe smartest way to use Polymarket right now isn't harder research. It's automated research. Most people try to keep up with Polymarket manually.

    2 min readAI

  32. AI in Customer Support, Education, and DataOpenAI has just launched "ChatGPT for Teachers," a new tool specifically designed for K-12 educators. This initiative is free for U.S.

    1 min readAI

  33. AI in Strategy, Product, and DataHad a great session with Ariel Michaeli of Appfigures on ASO. Here are some nuggets (best for last 😜): Screenshots that convert - Show real UI immediately...

    2 min readAI

  34. How far along are you with AI in your business - Post 1I'm curious where most of us stand in 2025... are you already saving hours every week with AI, just dipping your toes in, or still on the fence?

    1 min readAI

  35. How far along are you with AI in your business - Post 2I'm curious where most of us stand in 2025... are you already saving hours every week with AI, just dipping your toes in, or still on the fence?

    1 min readAI

  36. Google just yeeted Antigravity into orbitGoogle just yeeted Antigravity into orbit — an AI coding IDE that’s gunning straight for Cursor's throne, zero subtlety.

    2 min readAI

  37. There's a prevailing belief that AI, especially large language modelsThere's a prevailing belief that AI, especially large language models, will lead to widespread job loss.

    1 min readAI

  38. Engineering and Sales with AIApple just launched a Mini Apps Partner Program with a 15% commission on in-app purchases. This is a game changer for developers.

    1 min readAI

  39. The arrogance testThe arrogance test. Show an engineer AI-generated code and see if their immediate reaction is dismissal.

    1 min readAI

  40. AI in Engineering, Startups, and EducationThis is systemic influence, by design. OpenAI is quietly becoming the backbone of the hashtag#AI economy.

    1 min readAI

  41. AI in Strategy, Product, and StartupsMeanwhile in tennis world, something wild is happening. ATP world #1000 is closer to #3 than #3 is to #2. That is what winner-take-most looks like.

    1 min readAI

  42. If your advisor only knows your industryIf your advisor only knows your industry, you are paying for the same blind spots. The best second opinion often comes from outside the room.

    1 min readAI

  43. When Charles Schwab launched discount brokerage in the 1970sWhen Charles Schwab launched discount brokerage in the 1970s, Wall Street called it dangerous.

    1 min readAI

  44. Strategy and Data with AIToday I tried something that sounded impossible: building a web app with zero application code. No routes, no controllers, no logic files...

    2 min readAI

  45. AI and LeadershipA CEO plays word games to influence market bets and it shines light on the vulnerability of betting platforms.

    1 min readAI

  46. AI in Strategy, Leadership, and StartupsMost people think taste is innate. In reality, it's a skill you earn through tinkering. You don't develop good taste by consuming more content.

    1 min readAI

  47. Something smells fishySomething smells fishy... 🤷‍♂️ Is Flippa.com just trying to get me to sign their "Manager Agreement" (ie they take a cut even if I find a different buyer)?

    1 min readAI

  48. AI in Operations, Product, and EngineeringYou've probably heard the hype: "AI agents will replace entire workflows." Yet less than 5% of them actually work in production.

    1 min readAI

  49. AI, Operations, and StartupsEveryone in tech talks about AI agents for crypto trading, content generation, or writing code.

    2 min readAI

  50. AI in Strategy, Engineering, and HiringIf your team always defaults to React, your business is silently paying a tax on performance, innovation, and opportunity.

    2 min readAI

  51. ChatGPT remembers everything aboutChatGPT remembers everything about you. Claude barely remembers at all. That single difference explains why the two products now feel worlds apart.

    1 min readAI

  52. AI in Automation, Strategy, and ProductChasing full autonomy for AI Agents is wrong. They need to be the scaffolding human expertise.

    1 min readAI

  53. AI in Strategy, Product, and EcommerceMost ecommerce leaders are still optimizing (and spending) for Google search. But your next "customer" won't even be human, it will be an AI agent.

    1 min readAI

  54. AI and EducationMost leaders obsess over efficiency. But efficiency only matters if you are solving the right problem. Too often, organizations “optimize” broken systems.

    1 min readAI

  55. Engineering and Digital Transformation with AIThat's not a problem, that's a mess. A problem has a clear cause and solution. A mess is interconnected, dynamic, and resists being solved in isolation.

    1 min readAI

  56. Want to go viral on LinkedIn right now?It's extremely easy. All you have to do is... Declare that you either DO or DO NOT use AI in your writing, then proclaim the other side is wrong.

    1 min readAI

  57. 💥 LAST MINUTE Workshop for Founders💥 LAST MINUTE Workshop for Founders! I got a cancellation for this Monday morning (Pacific Time). It's a 2-hour workshop with me.

    1 min readAI

  58. Most founders don't run out of ideasMost founders don't run out of ideas. They run out of time, money, and patience… …because they built too much of the wrong thing.

    1 min readAI

  59. "Let's just get the requirements right the first time." That's how millions"Let's just get the requirements right the first time." That's how millions are wasted. Most people can't describe what they want in a way that translates...

    1 min readAI

  60. Most dev teams obsess over "doing it rightMost dev teams obsess over "doing it right." But "done fast" often beats "done perfect." Customers aren't buying clean code or test coverage.

    1 min readAI

  61. AI, Strategy, and Engineering"Clean code" is not self-documenting when it comes to why it was written that way. Most software documentation explains what the code does.

    1 min readAI

  62. The truth? Nobody really knows.OpenAI's GPT‑5 just dropped and everyone wants to sound like they know what it means. Nobody really knows. Yes, GPT‑5 is faster, smarter, more flexible.

    1 min readAI

  63. Customer Support and Engineering with AI😱 I gave Claude access to my inbox. Here's what it had to say... Your communication style signals: "I respect your time, I understand the real issues, and...

    2 min readAI

  64. ChatGPT has become such a suck-upChatGPT has become such a suck-up. Everything I ask is a “Great question!” and anytime I push back, “I should have thought of that!

    2 min readAI

  65. It's not the code that kills your vision, it's the unclear decisionsIt's not the code that kills your vision, it's the unclear decisions, the slipping timeline, and the invoices you don't understand.

    2 min readAI

  66. AI in Product, Sales, and DeliveryFive figures on "dev work" and nothing to show. 🤦‍♂️ • A boutique gym chain spent 6 months and $72k on mockups and 'exploratory' dev sprints • No TestFlight...

    1 min readAI

  67. AI doesn't raise the ceiling, it raises the floorAI doesn't raise the ceiling, it raises the floor. It accelerates baseline capabilities across your team. A sales director can prep tailored pitches in minutes.

    1 min readAI

  68. AI won't crush your industryAI won't crush your industry. It will magnify the flaw you already ignore. Most leaders still debate whether AI is disruptive or sustaining.

    1 min readAI

  69. AI in Automation, Hiring, and EducationWe accidentally deleted the entry-level job. And we haven't replaced it with anything better.

    2 min readAI

  70. AI can generate code faster than ever, but if you're not steeringAI can generate code faster than ever, but if you're not steering, you're not shipping. I've tested what happens when you let AI do most of the coding.

    2 min readAI

  71. Startups and Data with AITraining AI on public web data isn't theft, it's fair use... but we need laws that say so explicitly.

    1 min readAI

  72. The company known for baby photos (Meta) could end up more valuable than AppleThe company known for baby photos (Meta) could end up more valuable than Apple, Microsoft, or NVIDIA.

    1 min readAI

  73. AI in Customer Support, Strategy, and EngineeringCEOs beg for a magic AI prompt. A chef can’t cook without ingredients; your model can’t think without context.

    1 min readAI

  74. Now is the time to (re)buildNow is the time to (re)build. Most companies overpay to rebuild digital products because they wait too long.

    1 min readAI

  75. Beauty is the outcome of coherence, proportion,Beauty is not arbitrary. Design is not how it looks. It's how it works. Too many companies hear "design matters" and start decorating.

    1 min readAI

  76. AI in Product, Ecommerce, and DataYour app doesn’t need more features. It needs fewer while delivering ROI faster. Across 50+ product audits, I see teams sink 60% of budget into functions...

    1 min readAI

  77. Rethinking AI (21)Annoying enough that Frontier Airlines forces you to use the app instead of being able to add the boarding pass to Apple Wallet but at least make it large...

    1 min readAI

  78. Burger King spent like a League Two club and got a Champions League returnBurger King spent like a League Two club and got a Champions League return. 1.2B impressions from a team nobody can find on a map.

    1 min readAI

  79. AI in Strategy, Startups, and DataAI used to build startups from scratch. Now it is buying the ones you forgot existed. VCs like Khosla Ventures are rolling up call centers, accounting...

    1 min readAI

  80. Your favorite AI unicorn just tripped over its own numbersYour favorite AI unicorn just tripped over its own numbers. Builder.ai went from $1.3B headline valuation to insolvency in 72 hours.

    1 min readAI

  81. Would you leave the lights off when customers walk in?Your UI is your storefront. Would you leave the lights off when customers walk in? We started with static pages, then crammed them onto phones, then...

    1 min readAI

  82. I'm increasingly hearing about these frustrations from Startup Founders whoI'm increasingly hearing about these frustrations from Startup Founders who reach out to me. They've tried to hack an MVP with AI tools.

    1 min readAI

  83. The hashtag#EU just dropped a preview of its 2025–26 Startup & Scaleup StrategyThe hashtag#EU just dropped a preview of its 2025–26 Startup & Scaleup Strategy. Some parts are solid. Others felt already old.

    1 min readAI

  84. Sales and Data with AIPOV: Trying to save $132K in fees cost you $268K in lost earnings 😅 2x more money lost than saved 👎 I said in my last post as well that web checkout...

    1 min readAI

  85. Are you still caught up in the hype of AI replacing employees?The narrative has evolved. First, AI promised to 10x your workforce; then, it aimed to replace them.

    1 min readAI