A subdued, but very fitting 'grand finale' - love it! Air of mystery, tables and chairs stacked to the wall, The Man going about his usual business, but somehow it seems to be exposing a deeper layer (that closed laptop!) ...
Can I go out on a limb for a moment, and audaciously propose that The Man is 'God' or the great Force Of the Universe, and The Six are angels carrying secret messages for Mankind?
Nah, The Man that's you of course, and you're opening the laptop to write it all down - still the Master, but of your own little Universe!
Well this one really left me satisfied ... but of course we're now all eager to know: Is there Life after The End? But at least we're looking forward to The Recap!
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
Hahaha, dear leob, I love your bold theory! But you'll have to let me keep a few secrets for another day or two. See you in the recap post, coming tomorrow or the day after. Don't miss it! š
So thatās it. Oh no!!! It's the final Strategems.
The thing Iāll keep is Markās index entry. He doesnāt rerun the script or delete the old output. He rewrites one line: āIntervals checked. Source unverified.ā Thatās what a correct record looks like when you stop working on something, and itās harder than it sounds.
Pās line underneath is going above my desk. āDonāt fill in what you donāt know just to make it look complete.ā
And then the postmortem. System status: normal. Next analysis: not scheduled. I am still waiting. As an engineer, a process reporting healthy with no work left is the quietest failure there is. Nothing alerts, the status is accurate, and something has simply ended without the system being told. Forty-two chapters of refusing to turn incomplete observations into complete states, and the last thing it does is sit with the question.
Also, the man finally opens the laptop. After every chapter of bringing the right drink without being asked and saying almost nothing. I want to read whateverās on there more than any record in the story.
Thank you for all of these. Iām genuinely a bit sad, which is odd to say about a series with this much paperwork in it. š See you at the recap.
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
The moment I clicked Publish I found myself feeling a strange mix of emotions.
Is the Thirty-Six Stratagems series really over? Maybe it is. Maybe it isn't.
I can't quite tell whether I'm excited, reluctant to let go, or simply sad. It feels like I've finally reached the finish line, but now that I'm here, I don't quite know what to say.
As usual, the next post will be a recap of this stage of the journey.
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
Hahaha thatās such a good joke! I also fly through posts quickly sometimes. The estimated read time is just based on total word count, it never accounts for fast readers like you š
Earlier chapters even showed 20āminute reads. I kept worrying I couldnāt tighten my writing and made them way too long!
Not a joke, seriously I had kept a stopwatch when I saw 11mins read but when i took a look, It was just 3 scrolls of my mouse scrollwheel, so I kept a stopwatch and to tell you exactly, I finsihed in 4min 32seconds.
I measure AI coding claims in public. LGTM ā Let's Get the Measurements: real runs, real numbers, receipts. Every article is an experiment I actually ran. Books & packs: aheadkit.gumroad.com
The line that stayed with me wasn't the walk-away, it was Mark rewriting his index note to "Intervals checked. Source unverified" instead of filling the gap in. I do the same thing in my own write-ups now. Anything I couldn't verify gets marked open rather than smoothed over, and it's harder than it looks, because an empty box on the page feels like unfinished work. Do you have a rule for how long a note like that is allowed to sit unverified before someone has to go chase the source, or does it just live there until it matters?
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
That's a really good question. I don't think every unverified note needs an expiration date, but I do think it needs a reason to be reopened.
My preference would be risk-driven rather than time-driven. If the missing source could affect a release decision, an acceptance result, or someone's next action, it needs to be investigated. If it has no practical impact right now, it can remain open, provided the uncertainty is clearly recorded.
But I'd also keep a trigger for reassessment. New evidence, a change in scope, or a decision that now depends on that information should bring it back into focus.
The important thing is not to let "unverified" quietly turn into "verified" just because enough time has passed.
An empty box isn't necessarily unfinished work. Sometimes it's the most accurate part of the report.
Thanks for raising this. I think Mark would have appreciated the question.
I measure AI coding claims in public. LGTM ā Let's Get the Measurements: real runs, real numbers, receipts. Every article is an experiment I actually ran. Books & packs: aheadkit.gumroad.com
Risk-driven makes more sense than a clock ā a note that can't change a decision doesn't need chasing, it just needs to stay visibly unverified. The line I keep is 'an empty box isn't unfinished work, it's the most accurate part of the report.' Do you write the reopen trigger into the note itself, like 'reopen if X', or does it live somewhere separate?
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
From a QA perspective, I'd keep it simple and close to the test or defect record.
If new evidence affects acceptance, release decisions, or regression scope, we reopen it. Otherwise, it stays clearly marked as unverified, with whatever evidence we already have.
We don't need a separate process for every unknown. We just need to make sure that "not verified" is never mistaken for "passed."
That's probably more useful in day-to-day testing than adding another layer of process. š
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
Mark rewriting his index entry to "Intervals checked. Source unverified." instead of rerunning the script is the detail I will remember from the whole series. Stopping honestly, with a record that says exactly how far the work got, is harder than either finishing or quietly deleting.
We learned a version of that the hard way at Ascendra Ventures while building memory for our own AI coding tools. A handwritten parser read every field one column off and produced numbers that looked fine and were all wrong. The rule we wrote down afterwards was that a confident wrong number is worse than a blank, so every extractor in the system has to be allowed to give up and say so. "Don't fill in what you don't know just to make it look complete" is the same rule, written much more nicely.
Now that all thirty-six are done, which stratagem do you think engineers ignore most often?
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
That's a great example, and I particularly like your rule: a confident wrong number is worse than a blank.
As for the stratagem engineers tend to overlook most often, I'd probably choose The Art of Strategic Withdrawal (čµ°äøŗäø).
Not because engineers should give up easily, but because we often confuse persistence with progress. We keep rerunning scripts, adding assumptions, and trying to force an answer out of incomplete evidence.
Sometimes the better engineering decision is to stop, preserve exactly what has been verified, document what remains unknown, and leave a clear path for whoever takes over.
Knowing when to continue is important. Knowing when continuing would only make things worse is just as important.
I suppose that's why I saved this one for the end of the series. š
Thanks for sharing your experience. Your parser example is almost a real-world version of Mark's final index entry.
I liked that the unresolved source stays unresolved even after the interval calculation checks out. Iād keep those as separate fields in an engineering handover too: what I verified, what I assumed, and what would justify reopening the investigation. Closing the laptop feels earned because thereās still a usable record, without pretending thereās a complete explanation.
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
Congrats on finishing the 36-part series, Lingfeng! As a 12-year-old founder building a PaaS on a phone, the line 'Donāt fill in what you donāt know just to make it look complete' hit me hard. Itās a great reminder to stay honest with our code and our users. Can't wait for the recap! āšÆ
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
Keep building, keep questioning, and don't feel pressured to make things look more complete than they really are.š
Looking forward to seeing what you build. See you at the recap! ā
Your 36-part series has been an absolute masterclass in thoughtful engineering and storytelling. I will definitely keep that advice close: no pressure to fake completeness.
Iāll keep building, keep questioning, and I canāt wait for the recap! See you there! šÆš
Hey, I'm Hema š
Developer, writer, and creator of Dev Opportunity Radar, a weekly series published every Friday on DEV, helping people discover opportunities they might otherwise miss.
We made it to #36!! And what a way to end the series, Lingfeng! I think this is a reminder many of us need. Sometimes we get so caught up in trying to make things work that we forget it's also okay to step back and walk away when it's time. Loved following this series. Thank you for taking us through all 36 stratagems š
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
Thank you so much for staying along for the whole journey! Itās bittersweet to reach the finale, and Iām glad the core message resonates with you š
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
I think so too. Knowing when to stop can be just as important as knowing when to begin. Sometimes walking away isn't giving up. It's recognizing what can no longer be honestly resolved, and leaving things in a state others can still understand.
I'm glad the quiet ending resonated with you. Thank you for reading!
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
JavaScript enthusiast with a growing passion for Python. I love exploring the world of programming, experimenting with code, and discovering new ways to turn ideas into reality.
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
Knowing when to walk away is underrated advice for engineering too: a lot of time goes into rescuing an approach that should have been dropped a week earlier. Enjoyed the framing.
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
Honestly, there are several more little details hidden in this final stratagem. Let's see if anyone can spot them all. Even the order of things has a meaning. Hahaha! š
AI testing storyteller. Writing the 36 Stratagems series ā ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA ā building AI test framework.
A subdued, but very fitting 'grand finale' - love it! Air of mystery, tables and chairs stacked to the wall, The Man going about his usual business, but somehow it seems to be exposing a deeper layer (that closed laptop!) ...
Can I go out on a limb for a moment, and audaciously propose that The Man is 'God' or the great Force Of the Universe, and The Six are angels carrying secret messages for Mankind?
Nah, The Man that's you of course, and you're opening the laptop to write it all down - still the Master, but of your own little Universe!
Well this one really left me satisfied ... but of course we're now all eager to know: Is there Life after The End? But at least we're looking forward to The Recap!
Hahaha, dear leob, I love your bold theory! But you'll have to let me keep a few secrets for another day or two. See you in the recap post, coming tomorrow or the day after. Don't miss it! š
Sure Thing I won't miss it ... !
So thatās it. Oh no!!! It's the final Strategems.
The thing Iāll keep is Markās index entry. He doesnāt rerun the script or delete the old output. He rewrites one line: āIntervals checked. Source unverified.ā Thatās what a correct record looks like when you stop working on something, and itās harder than it sounds.
Pās line underneath is going above my desk. āDonāt fill in what you donāt know just to make it look complete.ā
And then the postmortem. System status: normal. Next analysis: not scheduled. I am still waiting. As an engineer, a process reporting healthy with no work left is the quietest failure there is. Nothing alerts, the status is accurate, and something has simply ended without the system being told. Forty-two chapters of refusing to turn incomplete observations into complete states, and the last thing it does is sit with the question.
Also, the man finally opens the laptop. After every chapter of bringing the right drink without being asked and saying almost nothing. I want to read whateverās on there more than any record in the story.
Thank you for all of these. Iām genuinely a bit sad, which is odd to say about a series with this much paperwork in it. š See you at the recap.
Hahaha! Iāll keep whatās on that laptop screen a secret for now.š¤ See you in the recap post!š å¾åæ«ēļ¼ä¼°č®”ę天ęå天ļ¼ååå
The moment I clicked Publish I found myself feeling a strange mix of emotions.
Is the Thirty-Six Stratagems series really over? Maybe it is. Maybe it isn't.
I can't quite tell whether I'm excited, reluctant to let go, or simply sad. It feels like I've finally reached the finish line, but now that I'm here, I don't quite know what to say.
As usual, the next post will be a recap of this stage of the journey.
See you in the recap!š
Whatever, over or not, these are simply awesome! Just finished reading this. Niceš
Thanks so much! Itās a bit bittersweet to wrap this main arc up, and Iām glad this quiet ending landed well for you š
Wonder why it shows 11min read, I finished it in 3-5 mins

Maybe thats why my friends call me an "LLM" when it comes to reading
Hahaha thatās such a good joke! I also fly through posts quickly sometimes. The estimated read time is just based on total word count, it never accounts for fast readers like you š
Earlier chapters even showed 20āminute reads. I kept worrying I couldnāt tighten my writing and made them way too long!
Not a joke, seriously I had kept a stopwatch when I saw 11mins read but when i took a look, It was just 3 scrolls of my mouse scrollwheel, so I kept a stopwatch and to tell you exactly, I finsihed in 4min 32seconds.
The line that stayed with me wasn't the walk-away, it was Mark rewriting his index note to "Intervals checked. Source unverified" instead of filling the gap in. I do the same thing in my own write-ups now. Anything I couldn't verify gets marked open rather than smoothed over, and it's harder than it looks, because an empty box on the page feels like unfinished work. Do you have a rule for how long a note like that is allowed to sit unverified before someone has to go chase the source, or does it just live there until it matters?
That's a really good question. I don't think every unverified note needs an expiration date, but I do think it needs a reason to be reopened.
My preference would be risk-driven rather than time-driven. If the missing source could affect a release decision, an acceptance result, or someone's next action, it needs to be investigated. If it has no practical impact right now, it can remain open, provided the uncertainty is clearly recorded.
But I'd also keep a trigger for reassessment. New evidence, a change in scope, or a decision that now depends on that information should bring it back into focus.
The important thing is not to let "unverified" quietly turn into "verified" just because enough time has passed.
An empty box isn't necessarily unfinished work. Sometimes it's the most accurate part of the report.
Thanks for raising this. I think Mark would have appreciated the question.
Risk-driven makes more sense than a clock ā a note that can't change a decision doesn't need chasing, it just needs to stay visibly unverified. The line I keep is 'an empty box isn't unfinished work, it's the most accurate part of the report.' Do you write the reopen trigger into the note itself, like 'reopen if X', or does it live somewhere separate?
From a QA perspective, I'd keep it simple and close to the test or defect record.
If new evidence affects acceptance, release decisions, or regression scope, we reopen it. Otherwise, it stays clearly marked as unverified, with whatever evidence we already have.
We don't need a separate process for every unknown. We just need to make sure that "not verified" is never mistaken for "passed."
That's probably more useful in day-to-day testing than adding another layer of process. š
šę儽å„äøäøäøŖē³»åä»ä¹
åååļ¼äæåÆ
š
ęä»å¤©å°čÆå»å¾®äæ”å ¬ä¼å·é£č¾¹ 第äøę¬”åę ē»ę被亲ęēå°äŗ čæē»ę转åå°å®¶ęē¾¤äŗš
社ę»ļ¼ååå
åęÆē»ęęčµäŗ äŗ²ę
Mark rewriting his index entry to "Intervals checked. Source unverified." instead of rerunning the script is the detail I will remember from the whole series. Stopping honestly, with a record that says exactly how far the work got, is harder than either finishing or quietly deleting.
We learned a version of that the hard way at Ascendra Ventures while building memory for our own AI coding tools. A handwritten parser read every field one column off and produced numbers that looked fine and were all wrong. The rule we wrote down afterwards was that a confident wrong number is worse than a blank, so every extractor in the system has to be allowed to give up and say so. "Don't fill in what you don't know just to make it look complete" is the same rule, written much more nicely.
Now that all thirty-six are done, which stratagem do you think engineers ignore most often?
That's a great example, and I particularly like your rule: a confident wrong number is worse than a blank.
As for the stratagem engineers tend to overlook most often, I'd probably choose The Art of Strategic Withdrawal (čµ°äøŗäø).
Not because engineers should give up easily, but because we often confuse persistence with progress. We keep rerunning scripts, adding assumptions, and trying to force an answer out of incomplete evidence.
Sometimes the better engineering decision is to stop, preserve exactly what has been verified, document what remains unknown, and leave a clear path for whoever takes over.
Knowing when to continue is important. Knowing when continuing would only make things worse is just as important.
I suppose that's why I saved this one for the end of the series. š
Thanks for sharing your experience. Your parser example is almost a real-world version of Mark's final index entry.
I liked that the unresolved source stays unresolved even after the interval calculation checks out. Iād keep those as separate fields in an engineering handover too: what I verified, what I assumed, and what would justify reopening the investigation. Closing the laptop feels earned because thereās still a usable record, without pretending thereās a complete explanation.
I really like how you've separated those three things. Especially the last one: what would justify reopening the investigation.
An honest handover shouldn't just record what we know. It should also make clear what we don't know, and what new evidence would actually change that.
I think that's what Mark was trying to leave behind. Not a complete explanation, but a record that the next person can trust and continue from.
Thank you for reading so closely, Emily!š
Congrats on finishing the 36-part series, Lingfeng! As a 12-year-old founder building a PaaS on a phone, the line 'Donāt fill in what you donāt know just to make it look complete' hit me hard. Itās a great reminder to stay honest with our code and our users. Can't wait for the recap! āšÆ
Keep building, keep questioning, and don't feel pressured to make things look more complete than they really are.š
Looking forward to seeing what you build. See you at the recap! ā
Thank you so much, Lingfeng! ā
Your 36-part series has been an absolute masterclass in thoughtful engineering and storytelling. I will definitely keep that advice close: no pressure to fake completeness.
Iāll keep building, keep questioning, and I canāt wait for the recap! See you there! šÆš
We made it to #36!! And what a way to end the series, Lingfeng! I think this is a reminder many of us need. Sometimes we get so caught up in trying to make things work that we forget it's also okay to step back and walk away when it's time. Loved following this series. Thank you for taking us through all 36 stratagems š
Thank you so much for staying along for the whole journey! Itās bittersweet to reach the finale, and Iām glad the core message resonates with you š
I think knowing when to walk away is the hardest part of building anything, and this quiet ending landed really well for me.
I think so too. Knowing when to stop can be just as important as knowing when to begin. Sometimes walking away isn't giving up. It's recognizing what can no longer be honestly resolved, and leaving things in a state others can still understand.
I'm glad the quiet ending resonated with you. Thank you for reading!
@star123 Did I remember it this time?š
Congrats bro!! āŗļø
šš
Knowing when to walk away is underrated advice for engineering too: a lot of time goes into rescuing an approach that should have been dropped a week earlier. Enjoyed the framing.
Great take! Happy the message lands for engineers š
P.S. oh, and they FINALLY finished their coffee, lol! Very fitting, and it's those little details which made #36 such a gem :D
Hahaha, you noticed! After 36 chapters, I figured they had kept that coffee waiting long enough. š Some things deserve a proper ending too!
Honestly, there are several more little details hidden in this final stratagem. Let's see if anyone can spot them all. Even the order of things has a meaning. Hahaha! š
This is such a beautiful reflection, and exactly the kind of thinking we're so often missing! Thank you, Lingfeng! šš„°
Thank you so much for your warm words! Iām really happy this quiet message resonates with you š