What we build next, and what I need from you

• The brain works. The content is solid: 1,679 wiki entries, 266 journal entries, 95 mini-sites, 23 skills.
• The boot file has to shrink. Today it loads in full at every session, about 112,000 words before you type anything.
• That is two thirds of the memory gone at message one. In July the conversation was rebuilt from a summary once every three and a half messages.
• The knowledge should be diluted into the skills, not loaded as one block. About 450 lines out of 5,000 belong to every session; the rest belongs to the job that needs it.
• Writing a file is not enough for it to come back. A skill has to point at it. Three of your last four remember-this requests produced a file nothing reloads.
• Office uploads are fixed. They were failing, fifteen separate times.

Your PST, right now

• Your whole archive is being digested by our own local model. It runs on our hardware, nothing is sent to anyone: fully private.
• It will distil a brain out of it. I give it to you read-only, so you can open it and see what seventeen years actually produced.
• Then we decide together how to join it to the brain you use today. Overwrite, merge, or something in between: I want your call, not mine.
• One boundary either way. The raw archive never goes into a brain. Only the distilled version does, and bank details and personal folders stay out.

• I am about to build brains for your colleagues. Five to ten to start with, plus a weekly session to find out where they fall short.
• Every one of them starts with a short audit of their own mailbox, the same way yours did. That is what tells us what each brain should be good at.
• Each person connects their own mailbox. One link, they authorise it themselves, nothing else to do and no archive to export.
• The question I need you to answer: how should the know-how of GR Recruitment reach those brains? Should each one answer in the house style, aware of GRAM and the rest of the method?
• GRAM and the method are strong. They are not a company brain, and I do not think one is what you need.
• What works is skills plus a well-structured project runway. The method becomes something each brain calls when the task needs it, rather than a second brain sitting above everyone.

• Two options. Option A, fix and optimise what is there, € 4,340. Option B, full rebuild with multilingual, € 5,700.
• Timing. About two months for Option A. Three to four for Option B, most of it spent agreeing the layout rather than building it.
• Running cost is the same either way, € 610 a year for hosting and maintenance, invoiced by GR WorkTech.
• The build is the same either way; what changes is what we build it with. Four routes, below.
• The build is invoiced by giobi.com, which is what makes it eligible for the subsidy.

Four ways to build it

• Framer. A design tool that also hosts the site. Quick to make it look good. The site lives inside their product, and what we can add later is limited to what they allow.
• WordPress and Elementor. What the site runs on today. Anyone can edit it by dragging blocks. It is heavy, it ages badly, and a brain reads it poorly because the content is buried in the page builder.
• AI-native static. Plain code, nothing underneath. Fast, cheap to host, and the brain can read and rewrite it directly. Content changes go through the brain or through me rather than a visual editor.
• Laravel, as an extension of the hub. The site becomes part of hub.grworktech.com: same login, same data, forms and candidate flows connected to what is already there. The most capable option, and the most work.

• What it is. The place where budgets, signatures and documents live, and where the brains can reach them. You are reading its budgets from it right now.
• Build € 1,900 one-off, budget here. Retainer € 100 a month for hosting, updates and support, invoiced by GR WorkTech.
• It grows with what you ask of it. What it does next is a budget question, not a technical one.
• Half an hour together is enough to decide what it has to do for Generations, and I would rather do that before building further.

Line by line, with the budget behind each one

If Generations wants to contribute actively to the R&D of the product, the amount can be decided between us; sixteen thousand euros is the figure I had in mind for what the product needs this year.

If you are not inclined to put money into it, I will not bring this to the table again.

• This one cannot go through the subsidies. A grant attaches to an invoice for work actually delivered, and seed money is not work.
• It also has to land in GR WorkTech, not in me. The waiver mechanism removes a cost from the company; it does not put cash in its account, and cash is the point here.
• So it comes in the plain way, as capital or as a shareholder advance, and it stays separate from everything in the previous chapter.