There exists a piece of technical advice so ubiquitous it’s become a cultural touchstone: “Have you tried turning it off and on again?” It’s the punchline of IT jokes and—perhaps surprisingly—one of the most profound insights into complex systems ever distilled into everyday language.
It feels like a dismissal. A deflection. A ritualistic incantation performed when actual understanding fails. And yet. And yet. It works. Disturbingly often. And not just for software.
Why Can’t I Just Roll The Boulder Back Up The Hill?
Complex systems exhibit several characteristics that make pinpointing and rectifying specific faults extraordinarily difficult, often rendering traditional debugging methods impractical or even impossible.
-
Accumulation of Invisible Internal States:
- Every interaction leaves subtle, persistent traces that aren’t fully documented or transparent, creating “hidden state” that makes current behavior a function of entire system history.
- Like sediment building up invisibly at the bottom of a lake, past inputs, minor events, and adjustments leave internal traces that aren’t obvious from the outside. The system’s behavior now depends on a history that is largely unrecorded and inaccessible, making diagnosis based only on the present state misleading.
-
Path Dependency:
- The fundamental structure and operational pathways of a complex system are often permanently constrained by early decisions or historical events, even if those reasons are long gone.
- Like a city built around ancient cow paths, the system is “locked-in” to certain ways of operating because subsequent developments were built upon these initial foundations.
-
Dense Interconnections & Non-Linearity:
- Components in complex systems are intricately linked, forming a dense web of connections and feedback loops.
- An action in one part doesn’t just affect its neighbors; it can trigger surprising and disproportionately large consequences in distant, seemingly unrelated areas. Small changes can amplify into massive shifts or cascades.
- This makes interventions risky, as fixing one issue might unexpectedly break something else due to these tightly coupled, non-proportional cause-and-effect chains.
-
Overwhelming State Space:
- As a system incorporates more parts, the sheer number of potential ways these parts could interact with each other explodes exponentially. Even with a moderate number of components, the possible combinations of states and interactions become astronomically vast.
- Pinpointing a specific faulty interaction or state combination becomes practically impossible because the search space (all the possibilities you’d need to check) is simply too enormous to explore exhaustively.
-
Opacity & Emergence:
- Many complex systems, especially natural or social ones, weren’t designed with a master blueprint; they evolved or self-organized.
- We often lack a complete, predictive understanding of their fundamental operating principles. Their most characteristic behaviors often emerge unexpectedly from the interactions of lower-level components, rather than being explicitly programmed.
- Debugging is hampered because there’s no definitive instruction manual or architectural diagram; we’re dealing with phenomena that arise organically and may not be fully explainable in a simple, top-down manner.
Wiping the Slate Clean
The reset strategy sidesteps the immense difficulties of precise diagnosis and repair by embracing a radically different approach: wholesale abandonment of the problematic state. Its effectiveness relies on:
-
Return to a Known, Simpler Baseline:
- The core value proposition of a reset is the replacement of the current, complex, error-ridden, and poorly understood state with a predefined initial state.
- This starting state is typically much simpler and, crucially, is known to be functional, at least at a basic level. Like restoring a computer to factory settings!
-
Breaking the Chains of Path Dependency:
- By reverting to an initial state, a reset effectively severs the links to the system’s specific historical trajectory. It escapes the suboptimal configurations and constraints imposed by past events, offering a chance to start fresh, potentially avoiding previous pitfalls (though there’s no guarantee it won’t fall into similar traps again).
-
Economic Rationality Under Uncertainty:
- In many situations, the expected cost (in time, effort, resources, and potential for further damage) of attempting to fully diagnose and precisely fix a problem in a highly complex, opaque system outweighs the costs associated with a reset (downtime, loss of current state).
- When diagnosis is difficult and expensive, and the outcome uncertain, choosing the reset is a no-brainer. It accepts a known, manageable cost (the reset disruption) to escape an unknown, potentially unbounded cost (fix one bug, two show up).
-
Drastic Simplification of State Space:
- A reset momentarily forces the system into a much simpler configuration, drastically reducing the number of variables and interactions that need to be considered. This makes the system’s immediate behavior more predictable and manageable, allowing core functions to be re-established before complexity inevitably begins to re-accumulate.
Applications Outside Software
Healthcare
-
Corticosteroid pulse therapy:
- When autoimmune diseases like multiple sclerosis or lupus flare up severely, patients receive extremely high doses through an IV for just 1-3 days. The massive steroid dose effectively shuts down the entire immune response temporarily, halting the attack and giving the body a chance to reset. After treatment, the immune system gradually returns but often with a different balance that’s less likely to attack the body’s own tissues.
-
Autologous stem cell transplant:
- Replaces a patient’s entire immune system. First, doctors collect stem cells from the patient’s own blood. Then they use strong chemotherapy to completely eliminate the existing immune system. Finally, they return the previously collected stem cells, which rebuild a new immune system from scratch. This approach has shown remarkable results for multiple sclerosis, with about half of patients seeing their disease progression completely stop.
-
Fecal Microbiota Transplantation:
- FMT transfers poop from a healthy donor into a patient’s intestinal tract, effectively introducing an entire balanced community of beneficial microorganisms at once. Originally developed to treat antibiotic-resistant C. diff infections, this approach is now being studied for inflammatory bowel diseases like ulcerative colitis. There are many anecdotal success stories for MS, CFS, IBD.
-
Organ transplants:
- Self-explanatory. Uninstall, reinstall.
-
Platelet-Rich Plasma (PRP):
- The procedure involves drawing a patient’s blood, processing it to concentrate platelets up to five times their normal level, and then injecting this concentration back into damaged areas. These platelets release growth factors that “stimulate or speed up the healing process, shortening healing time for injuries, decreasing pain and even encouraging hair growth.” Personally, I have a friend who healed a stubborn arm injury after this.
-
Intravenous Immunoglobulin (IVIG):
- Uses concentrated antibodies from thousands of blood donors to reset dysfunctional immune systems. By “preventing the body from attacking itself and decreasing several types of inflammation,” IVIG effectively treats conditions ranging from immune deficiencies to autoimmune disorders like dermatomyositis and lupus.
-
Psychedelics and Neural Annealing:
- Psychedelics may help treat conditions like depression and PTSD by temporarily disrupting rigid brain patterns and allowing healthier connections to form. Scientists describe this process as similar to “annealing” in metallurgy—heating a metal to make it more malleable before cooling it into a new shape. In the brain, psychedelics temporarily increase neural activity and flexibility, creating a window where entrenched thought patterns can be reshaped.
-
Fasting and Elimination Diets:
- During prolonged fasting (typically 3+ days), the body activates a cleanup process called autophagy, where cells break down damaged components and recycle them—essentially clearing out cellular junk and inflammation triggers. This process affects multiple systems simultaneously, resetting inflammation levels, immune function, and cellular signaling throughout the body. This system-wide reset can sometimes break cycles that medications cannot. Similarly, elimination diets work by removing all potential food triggers at once (often including dairy, gluten, soy, nuts, and other common allergens) and then systematically reintroducing them to identify problems.
-
Histone deacetylase (HDAC) inhibitors
Valproic acid in combination therapies to effectuation something like an epigenetic reset on methylation and acetylation patterns. HDACi’s keep chromatin in a more open, transcription‑permissive state (where deacetylation is essentially a nuclear response to environmental states that accumulates with time to close chromatin to certain transcription options; this is narrowly adaptive but ultimately difficult to reverse).
- Often used in cancer.
Other Domains
Organizational design:
- Easier to create a new innovation department within a monolith than to try to change the entire org.
Educational unlearning:
- Sometimes teaching a concept from scratch is more effective than trying to correct ingrained misunderstandings (see: negative transfers). Famous Max Planck quote:
“A new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it.”
Economic policy:
- Market cycles that permit controlled failure and renewal outperform those attempting to prevent all failures through intervention.
- Bankruptcy.
- Zero-based budgeting: Building budgets from scratch each cycle rather than iterating on previous budgets
If you have more examples, I’d like to hear about them, do share!
The Price of Starting Over
To be clear,I’m not advocating for reinventing the wheel (see Chesterson’s Fence). Nevertheless, the reset is a blunt instrument, and its application comes with significant drawbacks and considerations that must be carefully weighed:
-
Loss of Valuable Information and Adaptation:
- The reset is non-selective— the state being discarded may contain value that is difficult or impossible to recreate.
- Rebooting a computer loses unsaved work. Organizational restructuring can destroy valuable informal networks and institutional memory. Ending a long-term relationship erases shared history and learned compromises.
-
Significant Transition Costs and Disruption:
- The process of resetting itself is often disruptive and costly.
- Systems typically experience downtime during a reset, interrupting service or productivity. The transition period might involve significant effort, resource expenditure, and potential instability. Reorganizing a company, for instance, incurs direct costs and a period of reduced efficiency and confusion.
-
Failure to Address/Masking Underlying Flaws :
- A reset often addresses the symptoms of dysfunction (the system crashing or behaving erratically) without fixing the root cause. If the underlying design is flawed, or the operating environment consistently pushes the system towards failure states, resets will only provide temporary relief.
- Frequent reliance on resets can become a crutch, preventing the development of better diagnostic tools, more resilient system designs, or a deeper understanding of the system’s dynamics. Constantly needing to “reboot” an organization or a project suggests deeper issues are being ignored in favor of temporary fixes.
- Think of repeatedly rebooting software with a fundamental memory leak instead of patching the leak.
-
Risk of Reset Failure or Worse Outcomes:
- While the goal is to return to a known good state, the reset process itself can fail or even lead to a worse state.
- A botched operating system reinstall can render a computer unusable. A political revolution aimed at resetting a corrupt system can lead to civil war or a more oppressive regime. The transition is not guaranteed to be smooth or successful.
Conclusion
Hopefully I have made my case to embrace resets, or at least not to be embarrassed about it. Just remember to ask these questions:
- System legibility: How thoroughly can we understand the system’s internal state?
- Interdependence density: How tightly coupled are the system components?
- Historical accumulation: How much path dependency constrains current options?
- Interface preservation: Can critical external connections be maintained during reset?
- Modularity: Can portions be reset without total system disruption?
And most importantly:
Did the author just make all that up?
Maybe. Write your own damn article (Just kidding - please share any mistakes you find!)
Thanks to Jimmy for his thoughts & examples, & ultimape whose FMT mad-science experiment partly inspired me to write this post