We Don't Sell Saddles: Positioning With a Moat, and Without One

In November 2012, Stewart Butterfield shut down a game. Glitch — a whimsical, browser-based massively multiplayer world — died after roughly $17 million had been burned building it, and the company had almost nothing left except an internal chat tool the team had built to coordinate their own work. This is one of the most retold pivots in software, and the retelling usually skips the part that interests me most: the tool they had was, descriptively, a group chat product. And group chat, in 2012, was a dead category. IRC was decades old. Campfire and HipChat existed. Nobody was writing checks because the world lacked chat rooms. If Butterfield’s team had positioned honestly-by-default — “we make workplace chat” — they’d have been a slightly nicer entry in a category the market had already priced at approximately zero.

They didn’t. And the artifact that shows the choice being made deliberately, in writing, before launch, is a July 2013 internal memo Butterfield later published on Medium: “We Don’t Sell Saddles Here.”

The memo: choosing the shelf before the market chooses it for you

The saddles memo is the cleanest primary document I know of a company doing, on purpose, the thing I described in the positioning post — deliberately choosing the context that makes your strengths obvious. Butterfield’s framing runs through an analogy: a company that sells saddles can position itself as a seller of saddles, competing on stitching and price against every other saddle maker — or it can sell horseback riding, taking on the harder but far larger job of growing the thing that makes saddles matter at all. His instruction to the team was to do the equivalent: don’t sell the chat tool. Sell the transformation — a different, calmer, more transparent way for a team to work — and accept the obligation that comes with it, which is that you now have to actually deliver that transformation, not just the software.

Notice what this move does mechanically. Positioned as group chat, Slack inherits the category’s competitive set (IRC, HipChat), its expectations (a utility, probably free), and its price anchor (low). Positioned against email — the actual thing teams would keep doing if Slack didn’t exist — it inherits a different comparison entirely: the universally resented default of office communication. Against IRC, Slack is a nicer client. Against email, Slack is relief. Same product. Same code. The choice of alternative did the work, which is exactly why April Dunford puts competitive alternatives first in her sequence and market category last: the shelf is the output of positioning, not the input.

The market’s verdict came fast. When Slack opened its preview in August 2013, 8,000 invitation requests came in on the first day — by Butterfield’s own account, 15,000 within two weeks. The long arc ran to one of the largest software acquisitions ever: Salesforce bought Slack for $27.7 billion in December 2020. I’m always careful with outcome numbers this large — plenty of things other than the memo happened in those seven years — but the narrow claim survives any discount you apply: the position was chosen before the traction, not narrated after it.

What the position was anchored to

Here’s the question this post actually exists to ask, and it’s one the Slack story alone can’t surface: what was the “against email” position anchored to? When you claim a position, you’re claiming a differentiation — some reason the customer should file you on the flattering shelf rather than the default one. That differentiation has to live somewhere. And where it lives determines whether the position is an asset or a liability.

Slack’s differentiation lived in things Slack controlled. The product experience — searchable history, integrations, the feel of the thing — was theirs to build and theirs to compound. The “way of working” the memo promised to sell was theirs to define and evangelize. Competitors could copy features (Microsoft eventually did, with force), but nobody could revoke Slack’s differentiation by changing an API price, because the differentiation wasn’t rented. The position was ambitious, and it obligated the company to enormous ongoing work — the memo is explicit that selling horseback riding is harder than selling saddles — but the ground under it was theirs.

That’s the property to keep your eye on, because the next company had everything Slack had — a sharp position, a real market, spectacular growth — except that one.

Jasper: five weeks of runway on a borrowed moat

Jasper was, by the visible numbers, one of the fastest-growing software companies of its era. It positioned itself squarely and successfully as an AI writing assistant: marketers and creators paid real subscription money for AI-generated copy at a moment when that capability felt like magic. In October 2022, Jasper raised $125 million at a $1.5 billion valuation on the strength of that position.

On November 30, 2022 — five weeks later — OpenAI released ChatGPT, and gave the core of Jasper’s value proposition away for free.

Sit with the mechanics rather than the drama, because the mechanics are the lesson. Jasper’s position — the AI writing assistant — was real, well-executed, and commercially validated. But the differentiation underneath it was, substantially, privileged early access to GPT-3 wrapped in workflow and templates. The magic in the demo belonged to a model Jasper didn’t own, built by a supplier Jasper didn’t control, who was free at any moment to sell that magic directly — or hand it out free as a research preview. When that happened, Jasper’s shelf collapsed in the most brutal way a shelf can collapse: not because a competitor beat them at their category, but because the category’s core value went to a price of zero. “AI writing assistant, $49 a month” is not a position that survives “AI writing assistant, free, from the people who make the AI.”

The consequences arrived through 2023: Jasper reportedly cut its ARR forecast by 30 percent or more, ran layoffs in July, saw its founders step out of day-to-day leadership in September, and cut its own internal valuation by roughly 20 percent. (You’ll find a “revenue dropped 54%” figure circulating in aggregator posts; I haven’t found a credible source for it, so I’m not using it — the documented forecast cut and valuation cut are damning enough.) The company then did the only available move: it repositioned, away from “writing assistant for anyone with a blank page” toward an enterprise marketing copilot — workflows, brand voice, campaign operations — a position anchored in things a foundation-model vendor doesn’t automatically get for free: proximity to marketing teams’ actual processes, integrations, and organizational knowledge. Whether that reposition fully works is still an open question. That it was forced is not.

Positioning is a choice about what you anchor to

Put the two stories side by side and the lesson stops being “position boldly,” which is the lesson the Slack story teaches when it’s told alone. Slack and Jasper both positioned boldly. Both refused the commodity framing available to them. Both grew explosively on the strength of the chosen position. The difference is entirely in what the differentiation was anchored to — and that makes anchoring a strategy question, not a messaging one. It’s the same test I use for telling a real strategy from a wish: a coherent position, like a coherent strategy, has to rest on an advantage you can actually defend, and “our supplier likes us right now” is not one.

The way I’ve come to phrase it: anchoring your differentiation to something you don’t control is borrowing. Sometimes borrowing is the right call — Jasper borrowed GPT-3’s magic and converted it into two years of extraordinary growth and a nine-figure war chest, which is a better outcome than most companies that never borrow anything. But borrowed differentiation carries a repayment schedule you don’t set. The lender can call the loan whenever it suits them, and the more valuable the thing you borrowed, the more certain it is that they eventually will. Every “GPT wrapper” debate of the past few years is this exact question wearing new clothes: not “is it a wrapper?” but “when the underlying capability becomes free or ubiquitous, what’s left that’s yours?” For Slack, the answer was: the product, the network, the way of working. For Jasper in November 2022, the honest answer was: templates, and a head start measured in weeks.

There’s a third variant worth a sentence, because it completes the picture: sometimes the fix for a broken position is choosing a different shelf for the same asset. Notion, per Ivan Zhao’s telling, nearly died in 2015 positioned as a tool for building apps without code; the team retreated to Kyoto — reportedly on a $150,000 loan from Zhao’s mother, a detail that exists only in founder lore, so hold it loosely — rebuilt, and relaunched in 2016 positioned as a document tool people could adopt in five minutes. By 2025 that repositioned product had passed $500 million in ARR and 100 million users, with an $11 billion tender offer. The capability barely changed between the two positions. The anchor did: from “the ambitious thing we wanted to sell” to “the familiar thing users could actually file on a shelf.” Positioning killed the first Notion and made the second one — the same lever, pulled twice.

Put it to work

1. Write down what your differentiation is anchored to, and mark each anchor owned or borrowed. For every reason a customer picks you over the alternative, ask: do we control this, or does a supplier, a platform, or a partner? A model API, an app-store policy, a data feed, a channel relationship — each borrowed anchor is a loan, so name the lender and ask what happens to your position the day they compete with you or change the price. If the honest answer is “we’d have nothing left but UI,” you’ve found your Jasper exposure before the market finds it for you.
2. Position against the default behavior, not the category incumbent. Slack’s move — email, not HipChat — generalizes: ask what your best customers would actually go back to if you vanished, and write your positioning against that. If you’re positioned against a category peer while your real alternative is a spreadsheet, an inbox, or doing nothing, you’re fighting a war your customers aren’t in.
3. Run the free-tomorrow test once a quarter. Assume the most magical ingredient in your demo becomes free, universal, and vendor-supplied within twelve months — for anything built on foundation models this is less a stress test than a weather forecast. List what still differentiates you that day: workflow depth, proprietary data, network, distribution, trust. That surviving list is your actual position; whatever evaporates was borrowed, and the time to build the surviving list is while the loan is still cheap.

Further reading

• Stewart Butterfield, “We Don’t Sell Saddles Here” (Medium, 2014; written July 2013) — the internal memo itself, and one of the few positioning artifacts you can read exactly as the team read it before launch.
• TechCrunch, “The Slack origin story” (2019) — the fuller arc from Glitch’s shutdown through the pivot, with the beta-launch numbers in context.
• Contrary Research’s company memo on Jasper — the most careful public account of the rise, the ChatGPT shock, and the enterprise repositioning.
• Maginative, “Jasper cuts internal valuation as AI growth slows” (2023) — the primary reporting on the forecast cut and the ~20% internal markdown.
• April Dunford, Obviously Awesome — the framework both of these stories are illustrations of; competitive alternatives first, market category last.