German argues that post-2015 design patterns — autoplay, pull-to-refresh, streaks, variable rewards, short-form vertical video — are the dopaminergic equivalent of hydraulic fracturing: each technique extracts diminishing yield from a user whose baseline has been depleted by the prior one. The conventional 'deposits' (apps people open because they solve a problem) are tapped out, forcing the industry into increasingly aggressive extraction methods.
The editorial argues that ethics-based critiques of engagement design lose to quarterly metrics, but a structural resource-depletion argument eventually wins because the metrics themselves start to break — citing TikTok's falling watch-time-per-video as evidence the deposit is already showing signs of exhaustion. This reframes the conversation from 'we shouldn't' to 'we can't keep doing this.'
Commenters in the 165-reply thread challenge whether 'dopamine' is being used in a rigorous neurological sense or just as loose pop-neuroscience shorthand, and whether the extractive-industry framing is genuinely new or just a repackaging of existing humane-tech critiques from figures like Tristan Harris. The metaphor's vividness is doing more work than its scientific precision.
The post climbed past 300 points despite the chemistry quibbles because working designers and PMs reading the thread recognize the diminishing-returns pattern from their own product roadmaps — each new engagement technique yielding less than the last. The metaphor lands operationally even when readers reject it neurologically.
A blog post by Ilias German titled 'Dopamine Fracking' climbed to 349 points on Hacker News overnight. The piece is short, the metaphor is not: modern consumer software, German argues, is no longer engineering for user value — it's running an extractive operation on a shared neurological resource, and the easy reserves are gone.
The framing borrows directly from oil and gas. Conventional drilling — the equivalent of a product people open because it solves a problem — taps obvious, accessible deposits. When those deplete, the industry moves to hydraulic fracturing: high-pressure injection that fractures bedrock to squeeze marginal yield from rock that wouldn't otherwise produce. German's claim is that the entire post-2015 design vocabulary — autoplay, pull-to-refresh, streaks, variable-reward notifications, short-form vertical video — is the dopaminergic equivalent of pumping slurry into shale. Each technique extracts a smaller, more contested hit from a user whose baseline has already been pulled down by the last technique.
The HN thread (where the post is currently sitting) is doing the usual thing — debating whether the metaphor is too cute, whether 'dopamine' is being used in the loose pop-neuroscience sense, whether this is just Tristan Harris with better packaging. The reason the post broke 300 anyway is that the practitioners reading it recognize the operational pattern even when they quibble with the chemistry.
The usual frame for 'engagement design is bad' is ethical — *we shouldn't do this to users*. German's frame is structural: we *can't keep doing this* because the deposit is finite. That shift matters because ethics arguments lose to quarterly metrics, but resource-depletion arguments eventually win, because the metrics themselves start to break.
Look at where the curves are bending. TikTok's average session length is still climbing but watch-time-per-video has been falling for two years — the platform compensates by serving more, shorter clips. Instagram's Reels CTR is lower than Stories' was at the same point in the product's life. YouTube Shorts monetizes at a fraction of long-form CPM. Push notification open rates have collapsed below 2% across consumer apps despite send volumes increasing. Every one of these is the signature of an extractive industry: production volumes up, yield-per-unit down, marginal cost of extraction rising. That is fracking economics, not software economics.
The deeper point — and the one the HN commentariat is mostly missing — is what this does to the *measurement layer*. Engagement metrics were designed in an era when a user opening your app was a reasonable proxy for the user getting value from your app. Once you're injecting variable-reward schedules to drive the open, the correlation breaks. DAU goes up while NPS goes down. Time-in-app goes up while task completion goes down. The dashboards keep flashing green because the dashboards were built on the assumption that you weren't fracking. Retention as we measure it stops being a signal of product-market fit and becomes a signal of extraction efficiency.
Compare this to two adjacent patterns. Linear (which we covered earlier this week) is the inverse: it competes on raw user-perceived speed, which is a value primitive that can't be faked by injection — IndexedDB doesn't get more dopaminergic the more you ship of it. Slack, by contrast, is the textbook fracked field: notification volume per user has roughly doubled since 2019 while meaningful-message-per-notification has dropped by an order of magnitude, and the company is now spending engineering cycles on summarization features whose entire purpose is to clean up the slurry it pumped into its own aquifer.
The community reaction worth flagging is the disagreement between two camps in the thread. One camp — mostly ex-FAANG growth engineers — argues the metaphor breaks because dopamine is renewable; the brain isn't a finite reservoir, it's a regulatory system that adapts. The other camp — mostly indie devs and people who've left growth roles — argues that's exactly the point: the adaptation *is* the depletion. Tolerance is the equivalent of an aquifer dropping. Both are right, which is why this resonates.
If you ship consumer software, the operational implications are concrete and uncomfortable. Audit your engagement metrics for the gap between *extraction rate* (sessions, opens, time-in-app, notification CTR) and *value created* (task completion, retained intent, organic word-of-mouth, willingness-to-pay) — if the first set is up and the second set is flat or down, you are fracking. This is the cheapest diagnostic available and almost nobody runs it because the first set funds the team.
For B2B and developer tools, the lesson is to resist importing consumer patterns under the banner of 'engagement.' The number of dev tools that have added Duolingo-style streaks in the last 18 months is genuinely alarming, and they're being added by PMs who learned growth at companies whose underlying products were entertainment, not infrastructure. A streak on a CI/CD dashboard isn't a feature; it's a confession that the product doesn't pull users in on its own merits.
For anyone building anything with notifications: the per-notification yield curve is the canary. Open rates below 3%, dismissal rates above 70%, settings-page traffic correlated with notification volume — these are not 'optimization problems,' they're aquifer-depletion signals. The fix is not better targeting; the fix is sending fewer, larger notifications that carry actual information density, which is the software equivalent of switching from fracking back to conventional production. It pays out less per quarter and more per decade.
The useful prediction in German's framing is that fracked industries don't gently transition — they collapse on individual operators while the resource keeps producing under better-capitalized hands. Translated: the small consumer apps relying on dopamine tricks are going to see retention curves fall off cliffs over the next 18–24 months as user tolerance saturates, while the platforms with structural advantages (network effects, lock-in, identity graphs) will absorb the residual demand and look healthier than ever — right up until the same dynamic comes for them. The interesting builders to watch are the ones who looked at the curves a year ago and started building products whose primary metric is *intentional* use rather than incidental use. That's not a moral position. It's a bet on which deposits are left.
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.