I worked with Steve Jobs at Apple, where every OS update killed startups. AI founders are about to face the same thing

The article argues that AI is entering a new 'Sherlocking' cycle, where platform owners like Apple, OpenAI, Anthropic, Google, and Microsoft can absorb standalone features and pressure smaller startups. It highlights past examples such as Tile, Pebble, f.lux, Dropbox, Spotify, and 1Password to show that survival increasingly depends on deep enterprise integration, not just consumer-facing features. The piece is largely strategic commentary rather than event-driven news, so the market impact is limited.

Analysis

The market implication is not that large platforms are universally bullish; it’s that they create a selection effect. Horizontal AI leaders can commoditize single-feature startups quickly, but that mostly pressures consumer wrappers and light workflow tools rather than companies embedded in regulated or operationally complex environments. The second-order winner set is enterprise software vendors that sit inside approvals, compliance, identity, and data plumbing, because those layers are expensive to rip out and harder for foundation labs to replicate at scale.

The clearest public-market beneficiaries are DBX and SPOT, both of which have already survived platform encroachment by expanding into ecosystems rather than standalone functions. DBX still has room to compound if it keeps moving upmarket into collaboration and workflow infrastructure; the key variable is whether AI-assisted search/summarization becomes a feature or a product center. SPOT’s moat is less about audio playback and more about creator, discovery, and social graph depth, which makes it more resilient than consensus assumes if AI-generated media explodes, because distribution still accrues to the platform that owns attention and curation.

The caution is that AAPL, GOOGL, and MSFT are not short candidates on this thesis alone. The real risk is that these firms don’t need to win every vertical to compress startup multiples; they only need to bundle enough of the low-value surface area to slow adoption and force startups to spend more on services and integration. That means the time horizon for damage is months to years, not days, and the near-term trade is mostly in venture/private marks and small-cap software valuations, not in large-cap platform earnings.

Contrarian view: the memo’s premise is directionally right, but consensus may be underestimating how fast vertical AI companies can become sticky once they own process data and human-in-the-loop workflows. If a startup captures one mission-critical workflow and then expands adjacent modules, it can look like a feature until it becomes a system of record. That argues for preferring businesses with implementation revenue, high gross retention, and workflow expansion over pure model exposure.

Key Decisions for Investors

• Overweight DBX vs. a basket of AI workflow wrappers over the next 6-12 months; DBX has better downside protection if AI feature bundling accelerates, while wrappers face multiple compression first.
• Maintain a tactical long SPOT position into AI-generated media noise; use 3-6 month horizon and size for moderate upside, as ecosystem depth should protect retention better than headline risk suggests.
• Short public pure-play AI copilots / horizontal SaaS wrappers versus long AAPL or MSFT as a pair trade; the thesis is that platform bundling compresses standalone feature valuations before it dents platform earnings.
• Avoid outright shorting AAPL/GOOGL/MSFT on this theme; the better expression is long the disruptors that already embedded into enterprise workflows, since platform risk is a valuation headwind, not an earnings cliff.
• For venture/private exposure, favor companies selling compliance, identity, procurement, or data-integration layers; those names should command higher multiples because they benefit from AI adoption rather than being replaced by it.