When “One More Push” Isn’t Progress

That’s the stall—the moment you’re weighing when to stop a project. You feel it in your gut. Deadlines slide, code grows brittle, other projects stack up. When you’re deep in the weeds with a stubborn machine learning feature, every extra hour spent is an hour stolen from work that might actually move the needle somewhere else. It isn’t just about sunk cost. It’s about the opportunity cost you rack up while refusing to let go.

We’re wired to keep going. Every playbook in engineering and startups rewards the ones who don’t quit, who grind through discomfort, who treat every wall like an invitation to pivot or double down. We glorify consistency as the path to breakthroughs.

But here’s the thing. You can build a muscle for knowing when to stop. I needed a way to decide, fast—not more wishful tinkering. The rest of this piece is about running a weekly kill-or-keep review, setting clearer signals, and actually moving on—so your best energy goes toward work that still has somewhere to go.

What Keeps Us Grinding Past the Stall

Sunk-cost bias creeps in quietly, lacing itself around every hour, every late-night commit, every Slack thread defending the next experiment. Backtracking starts to feel bigger than changing direction—it feels like betraying not just the project, but the version of yourself who believed it mattered. The work becomes a point of pride, a marker of your place on the team. Some teams flip this logic entirely and actually reward cutting efforts early when experiments show it’s time to stop, making fail-fast the norm to avoid sunk cost. Most of us aren’t so lucky. I wasn’t protecting the project. I was protecting the me who started it.

Then there’s the double-edged fear. The first: looking like a quitter, losing face with your peers, or just telling yourself you didn’t see it through. The second: the dread that real progress could be one push away, and you’re about to miss it. Even now, part of me whispers, what if the next run is the one?

I like to look at it the way you’d watch training loss in a stubborn ML model. You burn compute, you rack up graphs, but when it’s plateaued, you’re just spending energy and not getting closer to the goal. It’s treadmill progress. You sweat, screens flicker, but the distance doesn’t change. Sure, incremental tweaks feel like movement, but the meter’s stuck.

What really traps you is the obligation loop. You start showing up because you said you would, not because new evidence says it’s worth another go. I’ve caught myself sticking to a project out of duty, not conviction. If you recognize that pattern, you’re not alone.

Consistency is valuable. But knowing when to let go? That’s wisdom. If compounding effort isn’t leading to compounding growth or learning, you’re not building. You’re just repeating the motions. The sooner you spot it, the easier it gets to move on and make space for the work that wants to move with you.

Sometimes, oddly enough, the friction is more obvious in unexpected places. A while back I spent hours debugging why a CI job wouldn’t run at 2am. I was juggling between timezone configs and daylight saving quirks—felt sure there was some tiny fix I was missing. By sunrise, nothing had changed, except my irritation with the whole setup. That’s how you realize the real problem isn’t the bug, it’s the attachment to fixing it just because it’s yours. Eventually, I deleted the whole job and nothing broke. I haven’t checked the exact logs since.

Five Signals for When to Stop a Project: Making Stopping a Skill, Not a Gamble

There’s a difference between discomfort that signals opportunity and the kind that just signals a dead end. The five signals below are my way of turning vague second thoughts into clear checkpoints—things I can actually review at the end of each week rather than stew about in the back of my mind. If you use these, quitting (or sticking)—whether to push through or reset—becomes a decision, not a guess.

First signal: No progress. You’ve put in the cycles, tried new angles, maybe even rewritten that critical function—yet the outcome metrics haven’t budged. When I’m brutally honest, I notice this pattern with ML more than anything else. I keep tweaking hyperparameters. I’ll swap optimizers, change batch size, retrain, and the learning curve is still as flat today as it was last week. It’s easy to lose weeks here telling yourself you’re “so close,” but data doesn’t lie. If the core result isn’t improving, more effort probably won’t change much.

Second signal: Opportunity cost decisions. Every hour you spend wrestling with a stuck project is an hour not spent where your effort actually moves things. It’s exactly like burning CPU cycles on a process that’s stuck. Background tasks pile up, the whole system slows, and higher-value jobs get starved. Look at your calendar. What else isn’t getting your energy because this stalled thing is soaking it up?

Third signal: Forcing it. You know this one—there’s a real sense of dread before opening the repo. Motivation is running on pure willpower, not on excitement or evidence that progress is possible. Ask yourself: Are you showing up because you believe in it or just because you should?

Signal four: Wouldn’t start today. If you could go back, armed with what you know now—the failures, the lack of traction, the shifting priorities—would you choose to launch this work again? If the honest answer is “no,” consider that your permission slip to move on.

Signal five: Survival mode. This one sneaks up on you. Suddenly, you’re no longer evolving the project—you’re babying it just enough to keep it alive, patching minor bugs, responding to little fires, but never charting a path to better. I realized I was nursing one of my own side projects, not actually growing it.

The roadmap was dead, the user feedback was recycled, and my only goal had become avoiding breaking anything new. It’s easy to slip into “just maintain,” especially when something used to show promise. But if everything is about upkeep and not momentum, it’s time to decide when to sunset features and free up the space—whether that means an orderly sunset, archiving, or handing it off to someone with new energy. Sometimes, the hardest part isn’t admitting you’ve reached survival mode. It’s believing you deserve to move on and let something else thrive with your attention.

How to Run Your Weekly Kill-or-Keep Review

Six months ago I started blocking off 30 minutes—every week, same slot. At first, this felt like yet another calendar ritual, but sticking to it changed how I navigated friction. The cadence matters more than the calendar invite.

Here’s what you do. For each active initiative, make a fast one-slide summary. Not a deck, not a doc—one slide. Spotlight the goals, the progress since last week, and what hasn’t budged. Pull up the five signals and check them. Is it stuck? Are you forcing it? Would you start it today? Are you just surviving, not building? The decision comes next—stop stalled initiatives, or choose to keep or hold.

If you hit “kill,” schedule your sunset steps. Code archived, docs closed out, team notified. If “keep,” tag a seven-day experiment—something measurable, small, and proving real movement. Use a framework for choosing to ship or refine so you balance speed and quality. Last step: take whatever hours you would’ve put into the stalled work, and commit (actually calendar) them to the highest-impact project you’ve got. Every hour spent forcing something that’s not working is an hour stolen from what could work.

Worried how it’ll look? Letting go isn’t quitting. It’s making space for something better. You’re reclaiming attention and focusing on what gives you leverage, not just what keeps your badge of perseverance shiny. Don’t skirt around the lesson. You made something, you learned, and now you get to redeploy that wisdom instead of defending sunk cost.

When you end a project, share the decision and the next bet—don’t just bury the change. When you say, “here’s why we’re shifting, here’s where our effort is going next,” you model that focus is courage, not cowardice, and rally the team after cancellation. That’s leadership.

Operationalize the reallocation within seven days. Block the newly freed hours on your calendar today and assign them to the initiative showing the most momentum and protect deep work time. Set one concrete, measurable checkpoint now, and bake in a review for next week. Lean teams use WSJF to always prioritize what matters most—quantifying cost of delay and sidestepping sunk costs by design. If you get clear on what’s moving—or what isn’t—you stay ahead of the bottlenecks and keep your best energy on real traction.

The whole ritual takes less time than a team standup but gives you back more than just hours. It gives you agency—one focused slot a week to trade false starts for actual progress. If you’ve been grinding past the stall, this is how you finally get out of the loop.

What Happens After: The Upside of Letting Go

Here’s the upside they don’t tell you about. The moment you stop forcing stalled work, everything else gets a whole lot lighter. Energy you were burning on “just one more fix” now goes to the things actually pulling for your attention—a feature customers keep asking about, the bug that’s throttling conversions, that interview loop promising insights no amount of code ever will. It’s the compounding principle in action: traction breeds motivation, progress starts to stack, and you finally feel like you’re building momentum, not just playing whack-a-mole. Success isn’t just about pushing. It’s about knowing when to stop a project. When you redirect attention to what has real pull, you give yourself leverage you never had before.

Looking back—like with that timezone thing—sometimes you drop an effort and life just moves on. You forget the drama around it almost instantly.

Last week, I stopped wrestling the stuck model. By the end of next week, the team shipped a new feature. Metrics moved today, not in theory.

Go further—bring your team (or just one friend) on board. Make it a ritual. Announce your weekly kill-or-keep move. Post it somewhere public, stick to it. If you want to anchor the shift, call it out—hashtag #YourMove. See what letting go can actually unlock, not just for you, but for everyone.

I won’t pretend this is perfect. I still get stuck hanging on longer than I should, especially when the work feels personal. Some projects just cling harder than others. Not sure I’ll ever clear that tension for good. But every time I let something go, that energy finds a better home.

If you’re watching for the signals, ready to cut loose where it counts, that’s real leadership. hashtag #Leadership.