Escaping Local Maxima: What Grief, High Jumping, and AI Taught Me About Transformation
A year after Margaret's death, I've learned something about transformation: sometimes moving forward means going backward first.
Grief taught me what gradient descent reveals mathematically: we can get trapped optimizing comfortable valleys instead of seeking higher peaks.
The question is whether we have the courage to look foolish while we figure out what those approaches might be.
Dick Fosbury revolutionized high jumping by looking ridiculous, jumping backward when everyone else was perfecting the straddle. He had something they didn't: foam mattresses that made new risks survivable.
AI is our foam mattress. It's not just a tool for doing our current jobs faster—it's a fundamental change in conditions that makes entirely new approaches possible.
The Mathematics of Moving Forward
Two weeks ago marked a year since Margaret died. I'd been dreading it for months. Not just the date itself, but the entire summer leading up to it. PTSD has a cruel way of time-stamping trauma. A year ago I was watching her fade. A year ago we were making impossible decisions. A year ago I still had her, even as I was losing her.
The body remembers what the mind tries to compartmentalize. July felt like walking through a ghost of itself. August was worse. By September, I was bracing for impact, certain the anniversary would flatten me.
It didn't.
Or rather, it did and it didn't. I felt the weight of it, but I also felt something unexpected: the faint outline of solid ground beneath my feet. Not healed - I'm not sure that's even the right goal - but different. Changed. Still learning how to live in this reality where she isn't.
Nora and I are finding our rhythm. Work has shape again. I can think about the future without it feeling like betrayal. I'm looking forward now, even as I carry the past with me. But here's what I keep bumping against: I don't know who I'm supposed to become in this new life.
The old frameworks don't fit. The person I was before doesn't exist anymore, but the person I'm becoming hasn't fully emerged. I'm somewhere in between, feeling my way forward in the dark, making small adjustments, hoping I'm headed in the right direction.
In my grief, I stumbled into an unexpected teacher: gradient descent, which I mentioned last year in my first very personal post about this new life. Not the mathematical concept I'd later come to understand, but the lived experience of it—the daily recalculation of which direction leads toward something resembling okay.
When you lose your North Star, your internal compass spins wildly. Every decision becomes a calculation with incomplete data. Do I get out of bed? Do I answer that email? Do I pretend today is normal or honor that it's not? Each choice, each small step, adjusts your trajectory by degrees you can't measure in the moment.
Machine learning algorithms use gradient descent to find optimal solutions. They take a step, measure the error, adjust, and step again. It's elegant in its persistence. Never a straight line to the answer, but a wandering path that feels its way downhill toward the lowest point in the loss landscape.
But here's what I didn't understand then, what the mathematics reveals with uncomfortable clarity: gradient descent can trap you.
You can find yourself in a comfortable valley, a local minimum where things feel stable enough. The algorithm (or the grieving human) settles there because every small step in any direction seems to lead upward, toward more pain, more uncertainty, more change. So you stop. You optimize for where you are. You convince yourself this is as good as it gets.
The technical term is a "local maximum" when you're trying to climb, a "local minimum" when you're trying to descend. Either way, you're stuck on a hill that isn't the highest peak or the lowest valley. It’s just the one you can see from where you're standing.
I found my footing again, eventually. But I found it by accepting something counterintuitive: sometimes moving forward means going backward first. Sometimes reaching the global optimum - the best possible version of your life, your work, your self -requires the courage to climb back out of your comfortable valley and descend into the unknown.
This isn't just about grief. It's about every moment we face transformation. And right now, with AI reshaping the landscape of work, we're all standing in our comfortable valleys, feeling the ground shift beneath us, wondering whether the path forward might require us to go down before we can climb higher.
The Man Who Looked Like He Was Falling
Dick Fosbury didn't set out to revolutionize high jumping. He set out to stop being terrible at it.
In high school, using the conventional straddle technique, the method every serious jumper used, the one refined over decades, and he couldn't clear 5 feet. His coaches worked with him. He practiced. He optimized his approach, his leg swing, his bar clearance. He was stuck at his local maximum, and it wasn't very high.
So he did something that looked absurd: he turned around and went over backward.
The first time people saw the Fosbury Flop, they thought he was injured, or confused, or both. One coach said he looked "like a guy falling off the back of a truck." Spectators laughed. Other athletes shook their heads. This wasn't innovation - it was a disaster in slow motion.
But Fosbury understood something profound: the straddle technique had reached its ceiling. Not for him personally, but for everyone. Athletes had been incrementally improving it for fifty years, gaining fractions of inches through minor adjustments. They were all trapped on the same hill, competing to see who could squeeze out one more centimeter from a fundamentally limited approach.
The difference? Fosbury was willing to go backward - quite literally - to find out what was possible on the other side of looking foolish.
By 1968, jumping backward over a bar, arching his back, landing on his shoulders, he won Olympic gold at 7 feet 4¼ inches. Within a decade, virtually every elite high jumper had abandoned the technique they'd spent years perfecting. Today, you can't find a competitive high jumper who doesn't use the Flop.
But here's what made Fosbury's revolution possible, the detail that matters more than his courage or creativity: the foam mattress.
Before foam landing pits, high jumpers landed in sand or sawdust. Landing on your back, on your neck, could break you. The straddle technique wasn't just tradition—it was survival. The optimization wasn't arbitrary; it was constrained by the landing conditions.
When foam mattresses arrived in the early 1960s, they didn't just make landing softer. They fundamentally changed what kinds of risks were survivable. They expanded the possibility space.
They made backward look feasible instead of suicidal.
Fosbury saw the changed conditions and asked a different question. Not "how do I get better at the straddle?" but "what if the straddle itself is the limitation?"
We're standing at a similar inflection point in our workplaces. AI isn't just a new tool for doing our jobs faster. It's the foam mattress. It’s a fundamental change in conditions that makes entirely new approaches possible.
The question is whether we have the courage to look like we're falling off the back of a truck while we figure out what those new approaches might be.
Because right now, most of us are still perfecting our straddle, measuring our progress in incremental improvements, competing on a hill that might not be the highest one available.
Our Current, Comfortable Straddle
If you work in an office, you've probably perfected your version of the straddle. Maybe it's the way you run meetings: the agenda template you've refined over years, the facilitation techniques that keep things moving, the follow-up email format that ensures accountability. You've optimized it. People compliment your meetings. You're good at this.
Or maybe it's how you manage projects: the spreadsheets, the status reports, the carefully calibrated check-ins that keep stakeholders informed without overwhelming them. You've found the sweet spot. You've climbed your hill.
Or perhaps it's your writing process, your client presentations, your strategic planning cycle. Whatever your craft, you've spent years (maybe decades if you’re like me) incrementally improving it. You've reached a level of competence that feels hard-won and valuable. You are not trapped. You are successful.
Except.
Except AI can now draft that follow-up email in seconds. It can generate meeting summaries, track action items, even suggest agenda optimizations based on team dynamics. It can create first-draft presentations, analyze data patterns you'd miss, simulate stakeholder responses to strategic options.
And here's where it gets uncomfortable: your first instinct is probably to use AI to do your current job faster. To optimize the straddle. To clear 5 feet 2 inches instead of 5 feet.
I did this too. When I first started working with AI during those grief-soaked months, I used it to handle the tasks that felt overwhelming: the emails I couldn't face writing, the research I couldn't concentrate on, the organizational work that required a functioning brain I temporarily didn't have. It was a coping mechanism, and it worked.
But using AI as a better mousetrap for catching the same mice misses the point entirely.
The real question isn't "how can AI help me do what I currently do?" It's "what becomes possible when the constraints I've been optimizing around disappear?"
When Fosbury stopped asking "how do I straddle better?" and started asking "what if I approached the bar completely differently?," he wasn't just improving his technique. He was escaping the local maximum that trapped every other jumper.
Going backward looks like this in practice:
It's the executive who stops using AI to write better memos and starts questioning whether memos are the right communication medium at all.
It's the teacher who moves past "AI can grade essays faster" to "what if assessment itself needs reimagining when AI can generate essays?"
It's the consultant who realizes that delivering polished PowerPoints faster isn't the value proposition anymore—but helping clients navigate profound uncertainty might be.
It's admitting that the expertise you've built, the processes you've perfected, the professional identity you've constructed might be optimized for conditions that no longer exist.
It is terrifying.
It means deliberately climbing out of your comfortable valley. It means a period where you're less competent, not more. Where you look like you're falling off the back of a truck. Where colleagues who are still perfecting their straddle seem more successful, at least for a while.
In my gradient descent through grief, I’ve learned that progress isn't linear. There are still days I when I’m Version 2.0, others I and I’m back to debugging the basics. But I’ve also learned something else: the courage to descend isn't about abandoning what you know. It's about recognizing when what you know has become a cage.
Margaret used to say "there is no wrong way." Not because all paths are equal, but because the fear of doing it wrong keeps us frozen on hills that aren't high enough (my trap). She approached each day as a prototype, not a final version. She understood intuitively what gradient descent teaches mathematically: you have to be willing to try, to fail, to adjust, to try again.
Our new foam mattress is here. Our new landing conditions have changed. The question is whether we're willing to turn around and jump backward, even though everyone's watching, even though we might look ridiculous, even though we've spent our entire careers learning to straddle.
Landing on New Ground
Here's what nobody tells you about finding a global maximum: you don't know you've found it until you look back and see how far you've climbed.
Fosbury didn't win gold at his first backward jump. He looked ridiculous for years. He failed often. He refined, adjusted, fell awkwardly, got up, tried again. The breakthrough didn't announce itself with trumpets… it emerged through iteration, through feedback loops, through the willingness to stay in the discomfort long enough to find the other side of it.
The same was true in my journey through loss. I didn't wake up one morning having "solved" grief. I woke up realizing I'd been adapting without noticing, that Nora's needs still pulls me into presence, that writing about AI has given me a framework for understanding my own transformation. The gradient descent had been happening all along, one clumsy step at a time.
This is what it means to be a work in progress: you don't arrive at finished. You arrive at different.
In our AI-transformed workplaces, "different" might look like:
The manager who realizes his value isn't in having answers anymore, but in asking questions AI can't formulate. Questions that begin with "what matters to us?" and "who do we want to become?"
The analyst who discovers that when AI handles pattern recognition, her real superpower emerges: explaining why patterns matter, translating data into story, helping people make meaning from information.
The creative professional who stops competing with AI's ability to generate content and instead focuses on what only humans can do: understand what's worth creating, judge what resonates with truth, know what the moment calls for.
These aren't small pivots. They're fundamental reorientations. They require us to release our grip on the identities we've constructed around our current skills and trust that new capabilities will emerge.
And here's the gift embedded in this uncomfortable transition: AI's limitations are revealing our superpowers.
Delivering Actual Value
Just as studying AI during my darkest months illuminated what makes us uniquely human - our capacity for meaning-making, our ability to love despite loss, our talent for finding poetry in chaos - watching AI attempt our work reveals the parts of our jobs that were never really about the tasks at all.
When AI can draft the document, the actual value was always in knowing which document the situation called for.
When AI can analyze the data, the actual value was always in understanding which questions matter.
When AI can generate the solution, the actual value was always in recognizing which problems are worth solving.
We've been optimizing efficiency for so long that we forgot to ask what efficiency is for. We've been perfecting technique without questioning whether we're jumping over the right bar.
Our new foam mattress - AI - doesn't just make new jumps possible. It forces us to reckon with what we're actually trying to accomplish.
Fosbury wasn't trying to straddle better. He was trying to get higher. When he released his attachment to the method, the height became possible.
We're not trying to send emails better, run meetings more efficiently, or produce reports faster. We're trying to create value, solve problems, connect with other humans, build things that matter. When we release our attachment to the current methods, no matter how well we've optimized them, new heights become possible.
Finding our Paths to Higher Ground
This requires something that no algorithm can replicate: the distinctly human courage to look foolish in service of something larger. The willingness to be a prototype when everyone around you appears to be a finished product. The capacity to find meaning in the descent, knowing that going downhill might be the path to higher ground.
Margaret understood this in her bones. "There is no wrong way" wasn't permission for carelessness (anyone who knew her knows that truth!). It was an invitation to experiment, to release ourselves from the paralysis of perfectionism, to recognize that being human means existing in the space between knowing and not knowing.
She approached each day as a prototype because she understood what AI is now teaching us at scale: growth is neither linear nor predictable. We're all learning to be better versions of ourselves, one interaction at a time, one awkward backward jump at a time.
The athletes who clung to the straddle technique didn't fail because they lacked talent or dedication. They failed because they couldn't see past their local maximum. They optimized what they knew instead of exploring what they didn't.
We stand at the same choice point. We can perfect our straddle, squeezing out incremental improvements while the world transforms around us. Or we can turn around, face the bar backward, and trust that the landing conditions have changed enough to catch us.
Nora, in her canine wisdom, has never once worried about looking graceful. She simply engages with each moment as it exists, not as she wished it to be. She can’t optimize… she can adapt. She can’t perfect… she can participate.
Maybe that's the real lesson here. Not that we need to abandon everything we know, but that we need to hold it lightly enough to discover what we don't know yet. Not that our current skills are worthless, but that they just might be components of something larger we haven't imagined.
The gradient descent continues. The path forward winds through valleys we can't see from here. But we've done hard things before. We've adapted to transformations we couldn't predict. We've found our way through landscapes that looked impossible to navigate.
We're works in progress, learning to jump backward, trusting that the foam will catch us, discovering that what looked like falling might actually be flying.
The bar is higher than we thought possible. But then again, none of us are who we were when we started jumping.
Resources from AIGG on your AI Journey
At AIGG, we understand that adopting AI isn’t just about the technology, it’s about people. People using technology responsibly, ethically, and with a focus on protecting privacy while building trust. We’ve been through business’ digital transformations before, and we’re here to guide you every step of the way.
No matter your type of organization, school district, government agency, nonprofit or business, our team of C-level expert guides - including attorneys, anthropologists, data scientists, and business leaders - can help you craft bespoke programs and practices that align with your goals and values. We’ll also equip you with the knowledge and tools to build your team’s literacy, your responsible practices, TOS review playbooks, guidelines, and guardrails as you leverage AI in your products and services.
Don’t leave your AI journey to chance.
Connect with us today for your AI adoption support, including AI Literacy training, AI pilot support, AI policy protection, risk mitigation strategies, and developing your O’Mind for scaling value. Schedule a bespoke workshop to ensure your organization makes AI work safely and advantageously for you.
Your next step is simple. Let’s talk together and start your journey towards safe, strategic AI adoption and deployment with AIGG.
