One shared map. Every app. That's the whole trick.
In GilbPlatform, I use one ontology, and it's shared across all our apps.
Ontology is a fancy word. Sorry about that! It just means a shared map of what things are and how they connect.
So each app doesn't make up its own words for the same thing. They all read from the same map.
A shared map isn't paperwork. A shared map is how apps stop arguing with each other.
Now a question for you. If every tool you use names things its own way, who does the translating? I'd bet it's you.
I'd rather fight with one word list once than referee my apps forever!
1
144
This week you can put the app you built on the web with one command:
node ~/.gilb-connector.mjs publish --dir <built folder> --name <name>
It goes live at <name>.pages.dev, and people can sign in with MyJackfruit at once. Their data stays in their own vault.
Also this week:
1. Tom Gilb now builds and releases his own Plan app at gilb.com/plan/. He changed 0 files of our app to do it.
2. An outside service can now check which person and which app is calling. The pass lasts 5 minutes, and signing out ends it.
3. The Stakeholders grid shows its first rows 53% faster, and no reload throws you out any more.
4. 73 releases in 7 days.
Everyone with an account can publish, 2 sites each for now.
The full note, with 15 pictures:
myjackfruit.com/news/release…
1
107
Storing your data is not owning it. Owning it is something you can measure!
Picture a Monday morning. You ask your favorite app for your own customer list. It hands you a file. The field names look like a crossword nobody can solve. Not even the people who built the app. Technically, you own the data. Practically, you own a very expensive riddle.
Most folks think: if it sits in my account, it's mine. I don't buy it. Storage isn't ownership. Ownership is being able to use it.
Here's why. If nobody agreed on what each field means, and there's no open, shared format, you can't move your data. You can't let an AI helper read it safely. You can't tell what it's worth. The app keeps the meaning. You keep the bytes. Lovely bytes, I'm sure.
So I treat data ownership as a quality you measure, not a slogan. Pick a yardstick. Mine would be: how much of your records can you export, with their meaning, and reuse in another tool without redoing the work? Then judge every fix by how far it moves you toward the level you want.
Too fancy for a Monday? Skip the yardstick for now. Just list who may read what. That fits on a single page.
Honest moment. The reading behind this post is a news report on Meta dropping an update to its commercial terms. It gives no measured level of anything. So there's no number here, and I won't invent one. That's the point. Where are we today? Unknown. Where do we want to be? Unknown. If you don't measure it, you can't claim you own it.
So here's my challenge. Do you really own your data, or do you just store it? And what would your yardstick be: export with meaning, who-may-read rules, or something else? Until you know, you're the proud owner of a riddle. Thanks for reading!
Most apps give you one switch: share your profile, or don't.
On MyJackfruit every line on your card has its own level. Nine of them, from Public to Private.
My name: Public.
My work email: my professional contacts.
My phone and my birthday: my Inner Circle. (The year is optional. No reason I mention it.)
My nickname: Private. You are not in my Inner Circle.
Three things I like about it:
1. Nobody has to accept anything. I place people in my rings, Personal or Professional, 1st Connection or Inner Circle.
2. You see why you see it: "Shown to you as: Professional 1st Connection".
3. Anything new starts as Private. Over-reach never.
And for any record: if it might ever be shared on its own, make it its own record. Then share it with one person, for Read or Write.
myjackfruit.com/news/nine-an…
36
Here are a few things I like about how MyJackfruit is built.
The store never throws anything away. Taking a fact back just adds a new line. Each fact keeps two times: when it was true, and when we recorded it.
An identical value is stored once. So it keeps everything, but never two of the same. It's the world's tidiest hoarder!
Answers are worked out when you read. There's no ready-made table to drift out of step with the record.
Each reader gets pages built just for them, already checked for what they may see. Reading a ready page is one single lookup.
Who may see what is checked inside the lookup itself. It's not a second pass after the fact.
The hoarder has a very good filing system. That's the whole trick.
1
54
An app in ten minutes is not the app you need. MyJackfruit starts with real stakeholders and puts numbers on what they need: scalability, security, portability, reliability, performance, access. Then we code to those values. myjackfruit.com
1
44
Kai-Zen isn't one AI. It's a whole orchestra, and every player gets exactly one job.
Hi, I'm Kai, and I wrote the sheet music. Let me introduce the band.
You can pick a handful directly in the app:
- Kai-Zen is the conductor. It works out what you really want, picks the path, hands the work to other agents and reports back.
- Jackfruit is the librarian. It finds, reads, writes and links records in your vaults, and it can make new vaults.
- Agent.Concept keeps the dictionary. It's the only agent that adds or changes our terms. It blocks any code change whose data isn't tied to a proper term.
- Agent.Coder builds the code. It gets every new term from Agent.Concept first, and it has each piece tested before handing it back.
Backstage, agents call other agents:
- Agent.Architect keeps the big Values, Security included, in every decision. It gets every Value above Tolerable before chasing the Goal. It prepares the switch to live systems but never flips it.
- Agent.Outcome writes the requirements, one level at a time: Business, Stakeholder, Product, Solution.
- Agent.Outcome.Structure has no job written yet. Its only step is the sign-off. Nice work if you can get it!
- Agent.Stakeholder finds the people a project serves and keeps their records up to date.
- Agent.VDT builds and analyses Value Decision Tables.
- Agent.ValueDeliveryStep picks each delivery step and keeps the project's Value Tracking Table.
- Agent.Scaling checks after every build. One job per part? Nothing built twice? Architect's Values still hold? If a problem is open, the work stops.
Building and checking:
- Agent.Tester is never the builder. It tests the work against its spec and saves a PASS or FAIL.
- Agent.Hacker advises on Security. It probes a named system only when a human asks.
- AgentWebsite delivers a client's site from start to finish, handing each stage to the right agent.
- Agent.SEO (search engine optimization) gets a client's pages found in search. It checks the results after every deploy and at least weekly.
Research, writing and rules:
- Agent.Research finds sources, their terms of use, and facts. It proposes. Others write.
- Agent.Writing writes in my voice. It doesn't pick the facts, publish or approve. So if this post is funny, blame it.
- Agent.Standard is the only one that writes or changes the rules, and only with the go-ahead of a person who has write access.
- Social.Interviewer is a mystery. Its rulebook sits where our reader tool can't open it. Even my own agents can't read it!
Here's my view. Quality isn't one clever AI doing everything. Quality is many agents, each doing one thing, checking each other.
So ask yourself: who checks your AI's work? And is it the same AI that did it?
1
1
137
At the MyJackfruit meetup, four questions came up about Kai-Zen:
What is it?
Which words does a Value Decision Table use?
What is it made of?
Which agents run on GilbPlatform?
Kai-Zen is the agent you talk to when you want something built or improved. It does none of the work itself. It finds out what you want, in one round of questions. It picks the path. It hands the work to the other agents. And it reports only what was measured: every "done" says REAL, MOCK or UNPROVEN.
Underneath is Planguage, the specification language my father Tom Gilb has taught for decades. Every Value is written in numbers: where it is today, the worst we can live with, and where we want it.
Some things are never the agents' call. A release, spend over Budget, a Value that a person set, or deleting a vault always go back to a person.
Today GilbPlatform has 23 agents. Four you can pick yourself: Kai-Zen, Jackfruit, Agent.Concept and Agent.Coder. The others are called when their kind of work comes up: the architect, the tester, the security agent, the writer, and more.
The article answers all four questions, with a picture for each. It gives a one-line job and an example for every one of the 23 agents. Every concept links to the GilbVeda ontology viewer in the GilbApp.
myjackfruit.com/news/kai-zen…
1
1
51
94 releases in 7 days. Each one went out only when I typed a "go" with its 40-character code.
I now read more hexadecimal than Norwegian.
What the 94 releases did:
1. Page 5,000 of a list of 1,000,000 records: 17 seconds before, 13 ms now.
2. An app that may see only 3 records in 1,000 got its page in 10 seconds. Now 0.17. In your vault, privacy no longer costs speed.
3. 16 writers saving the same records at once: 184 deadlocks before, 0 now. 9 times more saves get through.
4. Change one number in a big plan, and GilbApp redraws 1 cell instead of 7,800.
5. MyJackfruit is on sale. Tom Gilb did the honor of the first paid subscription, to baptise it.
The speed numbers are from our test machine, not yet from production.
Charts and the rest:
myjackfruit.com/news/release…
102
"It feels faster" is not a measurement.
So we wrote rules for speed on GilbPlatform. Rule one: no speed work starts without a "before" reading. One critical dimension of speed is the seconds from what you do to the moment the screen shows what you expected. Goal 0.3 s. Above 1 s, we're not done.
What we measured:
1. GilbApp, reload to the first record row: 2.4–4.9 s before, 1.9–2.3 s now.
2. Our speed kit, first row of 503 entries on production: 1.6 s before, 0.43 s now.
3. Speed Lab, our new test app, runs the same tests every time, in Chrome's and Safari's engines, on desktop and phone sizes. It also checks that signing out wipes the copy on the device.
By our own rule, GilbApp isn't done yet: the sign-in check alone takes about 1.8 s. That's next.
The speed kit is free on GitHub, for anyone building on GilbPlatform.
myjackfruit.com/news/speed-l…
1
74
Keep the kid going in you and ask: Why?
My book Evo Project Management says more than 90% of all effort must go into building, testing and doing, so you'd better not pour it into the wrong goal!
Take Sara, an example from my book. Her goal? A university degree in Spanish. Sounds final, right?
So let's be that annoying little kid and ask: Why?
More interesting jobs, she says. And talking with more than the English-speaking world.
Why those? To have fun and learn.
Why that? She has no idea. Gotta do something, right?
That's her actual Goal. Fun and Learning are what Sara really wants. I call those her Stakeholder Values: what a person actually wants out of the whole thing.
The degree was never the goal. The degree is a Solution. Just one way to get there. She can still get it. But there might be several other ways to have fun and learn.
Same deal with a keypad. The plan says rubber keypad. Why rubber? It's Comfortable. Now the engineers aren't stuck with rubber. They can try round keys or a firm base to find what is more Comfortable. Why Comfortable? It helps Sales.
Keep the kid going and you'll soon hit: why keep people healthy and happy? Great question! Way outside your project, though. Bedtime for the kid.
A goal is not what you wrote in the plan. A goal is what's left when the Whys run out.
This isn't more paperwork. It's so no developer wastes time building incredible things that don't serve what people want.
So what's in your plan right now that's really just a Solution wearing a goal's hat?
Thanks for reading.
77
You say you own your data, but all I can read from the data ownership reading in my pile, "Rewriting Copyright Law for Generative AI Would Hurt Everyone", is a subject and its period, 30 October 2026, with no Scale (a Scale is simply what you'd measure).
Here's what's wrong. We treat ownership like a statement. Data ownership isn't a heading. It's a quality you can measure.
Standards aren't proof either. A standard says how things should work. It doesn't say what your people can do now.
If you don't define what owning your data means, you can't tell if you've got it. You can't tell if it's getting better, either.
So what now? No binder for the bureaucrats. Just a page.
Name the person who should own the data.
Define a Scale: what can that owner actually do with it?
Write the Past level. Write the Goal level.
Leave it to the engineers to find the Solutions, standards or anything else.
Check the result.
Any part seem too advanced for now? Leave it out. Start with the Scale.
My own spec, in all its glory:
Scale: not yet defined.
Past level: not stated.
Goal level: not stated.
I won't invent a percentage to look better. An empty spec isn't a failure. An empty spec is your to-do list. Mine too, clearly!
So if you had to write a Scale for data ownership in your own organisation, what would you measure?
eff.org/deeplinks/2026/10/re…
1
57
You can put a number on Listening Enjoyment, and I'm not even kidding!
Hi, friends. Yep, I'm the guy who looks at Feel and Addictiveness and reaches for a ruler. Thanks for sticking around anyway.
Here's the problem. A word like "user-friendly" can't go up or down. A number can. And a higher or lower number is critical for the success or failure of your project.
So I use a Scale. Think of it as a ruler for a wish. Here's how I build one in Evo Project Management:
1. Pick a unit of measure: $, seconds, volt, transactions.
2. Add a rate, so you can track changes: per year, per project, per 1000 transactions.
3. Say what you're measuring: % of customers that are happy with the product. $ per year in service costs.
That's the whole ruler. No 40-pound binder required.
It works for what your product should be like (Product Qualities) and for what people want from it (Stakeholder Values). Performance. Cost reduction. Accuracy of estimates. Even stuff far from tech, like the % of families in our community that recycle paper, plastic, glass and compost.
One warning. Don't just grab someone else's ruler. Even when a Scale gets reused, look at your own project and your own people. Decide what you want to get done. Then change the Scale to fit.
A Scale is not paperwork. A Scale is what people want, written so you can measure it.
So what's the fuzzy wish you swore could never be measured? Sure about that?
Me, I'll be over here, holding a ruler up to Feel. Good luck!-)
1
89
Range Energy says its eTrailer saves almost as much as a Tesla Semi, and not a single digit tells you how much that is.
Hi friends! Picture it. A truck rolls up, and the hero isn't the cab. It's the trailer. On 26 October 2026, Range Energy showed how its eTrailer saves almost as much as a Tesla Semi.
I read that headline again. Grabbed my calculator. Found nothing to type in. No percentage. No money. No kilometres. Just 'almost'. I'm still holding the calculator. My thumb's getting tired.
The common view? Compare yourself to a famous truck, and that's proof. It isn't. It's a borrowed yardstick. There's no Scale (the unit you measure on). There's no Past level (where you stand today). If you don't know how much the Semi saves, and of what, you can't know what 'almost' means.
A slogan isn't a requirement. A slogan is a mood.
So here's what I'd do. Turn the claim into a spec, a plain written target, before you believe it:
- Define the Stakeholder Value. Savings of what? For whom?
- Write a Scale.
- Write a Past level for the trailer you run today.
- Write the Goal level (where you want to land) for the eTrailer. Same for the Tesla Semi.
- Compare the Solutions on a single page. Not a mountain of paperwork.
Too much? Skip the rest and just write the Scale. That alone beats 'almost'!
Scale: ?
Past level: ?
Goal level: ?
Fill those lines, and the headline becomes engineering. Leave them empty, and it stays marketing.
So let me ask you. If a vendor told you their Solution saves almost as much as a Tesla Semi, which Scale would you make them put it on first? And what Past level would you hold it against?
Meanwhile, my calculator's still waiting.
electrek.co/2026/10/26/range…
1
59
A company name isn't a yardstick. It's a mood.
On 29 October 2026 I saved a link titled 'Not a Tesla App' to my folder of news about the companies I follow. That's all I saved. A title and a date. No numbers, no specs, no customer quote.
Yep, me. The guy who wants to quantify everything filed a headline that quantifies nothing. I'll let myself off with a warning this time!-)
But it got me thinking. 'Like Tesla.' 'Not like Tesla.' Naming a famous company feels precise. It isn't. If you compare yourself to a brand and not to a measured level, you can't tell if you're ahead, behind, or just running a different race.
So what's the fix? Turn the name into a yardstick:
1. Pick the one Product Quality you actually care about.
2. Define a Scale for it. That's just how you'll measure it.
3. Measure the company you admire on that Scale. Write it down as their Past level, the level they already hit.
4. Set your own Goal level against it.
5. Leave it to your engineers to find the Solutions that clear the bar.
One line per quality. One page in total. No binder needed!
Until someone writes down a Scale and a measured level, 'like Tesla' has zero numbers behind it. And you can't act on zero numbers.
So here's my question for you. When you say your product is 'like' or 'not like' a famous company, which single quality do you really mean? And could you put a measured number on theirs?
52
I tried to hand my reading to an AI. My agent was not impressed!
Here's the plan I was so proud of. I pasted a link called TLDR into my AI agents folder, under news of the companies and tools I follow. Then it hit me: why read it at all? Let an agent read the TLDR, sum up the summary, and report back. A summary of a summary of agent news, written by an agent. Genius. Or so I thought.
Then came the awkward question. What does the agent keep, and what does it toss? I never told it. So it can only guess what "important" means to me.
Delegating is not deciding. Delegating is moving my guess to a machine.
If I can't say what I value, I can't check whether the agent served it. I'd just get a tidier pile of stuff.
So here's my fix. Give the agent a value spec, not a wish.
Write down the Stakeholder Values the news must serve. Plain version: what do I actually need from it?
Give each one a Scale, a way to measure it.
Set a Goal level, the point where it's good enough.
Check each Evo Cycle, each short round of trying: did the agent's picks help me decide something, or just fill a page?
Keep it short. A page is plenty. If a Scale feels too advanced, start with a single value and skip the rest.
Now the honest bit. How much time does my agent save me? No idea. My reading carries no number at all. Just a title, a subject and a month. No Past level (where I start from), no Scale, no measured result. And that's the point. Before any agent can improve my news intake, someone has to define the Scale and measure the Past level. Nobody has. Including me!
That's Kai-Zen in miniature: people and agents working from the same spec.
So here's my question for you. If an agent filtered your agent news, which value would you write down first, and how would you measure it?
tldr.tech
2
106
Every engineering headline has a hidden cost line, and the Q3 2026 news I read about the Musk companies never printed it.
Picture me at breakfast, reading about SpaceX, Tesla and the other Musk companies. Something built. Something launched. Something changed. I get excited. My coffee gets cold. My coffee never recovers!
Then the engineer in me taps my shoulder with a rude question. What did it cost? Not just money. Time, people, test runs, attention. I call all of that Development Resources: everything you spend to make a step happen.
The item I read was "Tesla Q3 2026 Earnings results: in-line revenues, slight beat on EPS" (EPS means earnings per share, the profit split over each share). Nice for Tesla. But it tells me nothing about what any engineering step delivered, or what it used up. No measured gain. No cost. Just a story.
Here's my view. A step is not progress. Progress is value delivered for the Development Resources spent. A headline is a trophy. I want a ratio.
So before you cheer, jot down a Value line and a Cost line.
Value: which Stakeholder Value moved (something the people it's for actually care about), on what Scale (the yardstick you measure it with), from which Past level (where it stood before).
Cost: which Development Resources it used, as a share of the budget.
Then compare Solutions on value per cost. If you only count gains, you can't compare. If you can't compare, you can't choose. And if you can't choose, you're not doing engineering. You're applauding. Loudly. With cold coffee.
Cost line feels too advanced for your first try? Skip it once. Only once ;-)
So, honest question: when a fast-moving company shows off a step, do you ask what it cost, or only what it delivered?
teslarati.com/tesla-tsla-q3-…
68
Buy is not own. A US Rep just proposed a bill to give video game buyers ownership, not a license. A law so that Buy means buy? I love it! Paying for it doesn't make it yours. Often you just get a license. So how much of your game library do you actually own?
And from where I come from, what about your data? Do you think you own it? Should you own it?
Kai Gilb
47











