Capture
Brain dumps land somewhere permanent instead of evaporating. A voice note, an end-of-day ramble, a thought at a red light. One inbox, given a shape on arrival so it's findable in November.
Kyle Johnston · GYST · running on myself five nights a week since July 2026
Get your stuff together.
On your terms.
GYST is the personal operating system I run on myself, five nights a week: capture what matters, hold a cadence, let the record compound, and keep the judgment human. It is a framework and a mental model, not an app. You own the files. You set the shape.
CAPTURE × CADENCE × COMPOUNDING × CURATIONCAPTURE × CADENCE × COMPOUNDING × CURATIONCAPTURE × CADENCE × COMPOUNDING × CURATION
01 / THE MENTAL MODEL
You don't have a goals problem. You have a follow-through problem.
Nobody quits in January because they picked the wrong goal. They quit in March because nothing held the thread on a week that went sideways, and by then there was no record of what had already worked.
The cadence is the first thing to go
The daily check, the weekly look-back, the monthly honest accounting. Every app assumes you'll do that part on your own. It's the part that quietly stops first.
Half your system lives in your head
What you're avoiding. What worked last time. Why you made that call in June. None of it is written down, so two busy weeks wipe it out.
Every app wants to be your system
And every one of them rents your memory back to you. Change tools and your history stays behind. A system you don't own isn't a system. It's a subscription.
02 / THE MECHANISM
Four layers. In order. Each one only works because the last one did.
Cadence
A daily check-in, a weekly review, a monthly reckoning, a quarterly checkpoint. The questions you'd ask yourself if you had the time and the distance. Short, blunt, and the same every time.
Compounding
The system reads itself back. Patterns surface. Rollups write themselves. Next month gets planned from what actually happened, not from what you remember happening.
Curation
The judgment layer. What to automate, what to protect, which new tool to ignore, and what your own data has been trying to tell you. This is the part a PDF can't do for you.
Where I come in
Deliberately boring. That's why it survives a bad week. The first three you run yourself. The fourth is where a human is still worth paying for.
03 / THE CADENCE, AS IT ACTUALLY RUNS
Four tiers. Each asks a different question.
These run as scheduled prompts against my own ledger, inside the tools I already use. Each tier reads the tier below it before it asks anything. A tier that re-asks the tier below it is dead weight.
Daily
The question
What moved, what blocked, and what's the one thing tomorrow?
What it does
A short retro at the end of the day, not a nudge at the start of one. The information only exists after the day happens.
The beats
- The day as it was.
- What moved. Specific, not generic.
- What blocked, and which kind of blocked.
- A money pulse, or silence.
- Tomorrow's one thing, with a day on it.
- A close chosen against the day.
Weekly
The question
What's only visible across days?
What it does
Reads the whole week back before asking anything. Sets three priorities, each with a named day, and an agreed hour when it needs someone else.
What it finds
- Stalls no single day reveals.
- Repeat blockers. Three of a kind is a system problem.
- Where the energy actually went.
- The wins you didn't count.
- One thing to stop, change, or kill.
Monthly
The question
Is the strategy working?
What it does
Reads the weeks, not the days. Days are the wrong fidelity for this question. It has to confirm the pace or amend the plan.
What it checks
- Pace against the real deadlines.
- Anything being built: has a real customer seen it?
- A product nobody has looked at is a hobby.
Quarterly
The question
Was the plan wrong?
What it does
Amends the plan in writing. Being wrong is expected: a plan built in July on July's information should change when October disagrees.
Five questions
- Did the gap close?
- Which assumptions broke?
- Are the goals still the right goals?
- What should be killed?
- The single biggest lever next quarter?
Four questions, asked on a schedule you don't have to remember.
04 / THE RULES THAT MAKE IT SURVIVE A BAD WEEK
Honesty mechanics, not motivation.
A rep has a floor.
Set a minimum so small you can't argue with it. One message to one person. Five dollars at the balance. Ten minutes of movement. A bad day should produce a rep instead of a zero. The habit is the asset; the size never was.
A zero day is data, not a failure.
Log it, tag why, move on. No moralizing, no streak language. A system that punishes you for reporting honestly gets lied to by the end of the month, and then it is measuring nothing.
Blocker type matters more than the blocker.
Avoidance, deprioritization, and a real constraint look identical in a log and need opposite responses. Three days stale: ask plainly which it is. Five days: make the call. There is no sixth day.
Buried wins get named.
Most people don't hide their wins, they misfile them. The real one shows up last, in a subordinate clause, as background for something else. The tell is position, not omission, so the review reads them back.
Probe before you reframe.
When someone renders a judgment on themselves — a rep on their quarter, a partner on their week, you on yourself — the first move is a question, not a counter-read. A confident reframe built on a fact you don't have isn't kindness.
The uncomfortable thing gets its own counter.
Building is work that has to happen, on dated hours with named deliverables. It never counts as the hard part. The ask, the invoice, the conversation you've been drafting for a week — count that separately, or the counter gets satisfied the easy way.
Some things are never graded.
Whatever you keep for its own sake — the family side, the hobby, the thing that makes a week feel like a life — gets logged and never scored. No streaks, no misses, no targets. The moment delight becomes a metric, it becomes a chore.
Personalize the close, never the read.
The mechanics above are the same for everybody. How a check-in ends is yours — a passage, a quote, a photo, a line to your kid. Mine closes with scripture chosen against the day that happened. Whatever you pick sits alongside an honest read, never on top of one.
None of it runs on willpower. That is the entire point.
05 / UNDER THE HOOD
One row per day is the spine. Everything else hangs off it.
Mine runs on Postgres and a folder of markdown, driven by scheduled prompts. Yours could run on Notion, a notebook, or a plain folder. The design principles are the product. The plumbing is a choice.
One source of truth.
Writes go to the store first. Readable files are regenerated in full from a query, never appended alongside. Appending creates two masters that silently drift.
Structured fields beat prose.
If a review has to read prose to find a pattern, that pattern-finding will silently stop happening. Anything a higher tier adds up gets its own field.
Fact and inference, marked apart.
Every row records whether you said it or the system inferred it. Otherwise, a month later, the system's guesses read back as reported history.
Sensitivity set at capture.
Normal, personal, restricted, decided when the row is written. Retrofitting it means reclassifying by hand, which means it never actually happens.
A registry of what's tracked.
Without one, the system can't tell “not tracked” from “tracked and missed.” Retire items, never delete them. Deleting destroys the history staleness runs on.
One contract file.
Every scheduled prompt reads the same shared schema and queries, and none of them invents its own. Four copies of the same logic is four chances to drift.
06 / PROOF, NOT DECKS
I'm not selling you a theory. I'm selling you the thing I use.
GYST started as my own second-half-of-the-year problem: debt to clear, weight to lose, a side practice to build, and a six-month window. The check-in fires five nights a week, I answer it, and the record keeps itself. I can screen-share it.
5 nights a weeksince July 2026
4 tiersdaily → weekly → monthly → quarterly
0 new appsit runs on files you already own
07 / BUILD IT WITH ME
Same system underneath. What changes is how much of the thinking you do alone.
For solo operators and small-team leaders in Katy, Fulshear, Richmond, and anywhere a screen-share reaches.
Map
Working session
We name your domains, your floors, and what "a day that moved" means for each. You leave with the shape of your system on one page, whether or not we build it together.
From $350 · credited to the build
Book the session →MOST PEOPLE START HERE
Build
Done with you
I build it with you, in your voice, on tools you already own: the capture inbox, the four scheduled check-ins, the ledger, the contract file. Then a tune once you've lived in it for a week, because week two is when it breaks.
From $1,500
Plan the build →Keep it honest
Monthly cadence + AI filter
A monthly review of what you actually used, what you avoided, what broke, and one system improvement. Plus the filter: every new AI tool that lands in your inbox gets a verdict. Adopt, ignore, or watch.
$250 / month
Start a conversation →You own the files. Nothing lives on my infrastructure. If we stop working together, nothing has to stop working.
Nothing autonomous talks to your clients. AI drafts and sorts. You decide, review, and own what goes out.
I don't hold your credentials or your data. Least privilege, revocable, and written down.
Starting points, not a menu of promises. Every one of these begins with what's actually in front of you.
08 / START
Tell me what's slipping.
Which version of this do you want: the session, the build, or the monthly? Say so, and tell me what's actually falling off the table right now. I'll read it.
Or find me on LinkedIn ↗