Ari’s Space*   est. 2026
Skip to content
ari*s space[ a. ]
← Libraryworkflows / 2026-09-13

The tool they wished they had

My journal through IBC. Built the tools this year on twelve years of experience. Nikon talks, friends on the floor, Subtrackt_ and CBA_ spinning in ASCII.

8 parts / Hover to explore. Click to read.

My journal through IBC. Built the tools this year on twelve years of experience. Nikon talks, friends on the floor, Subtrackt_ and CBA_ spinning in ASCII.

.neg -> .candy <-> .master
                     |
                     v
                   .share
                flat / one-way

.wrapper: alterations only
requires peer's .neg
LOCK / TASTE / MIX: .candy modes
Candy is a design account: the working and master paths retain science, while share deliberately flattens. A wrapper carries alterations only for a peer who already has the negatives; these names do not prove interoperability.
Context Pack (for agents)

Build the tool your industry has wished they'd had for decades.

I said that on the IBC floor this week. Partly to whoever was nearby. Partly because I needed to hear myself say it.

This is my September 2026 journal: Nikon talks, friends I get to hug when the industry puts us in the same city, notes between panels and coffee. Twelve years of experience came with me. The tools were built this year. Those are different clocks.

I built things I had wanted while doing the work. Something that keeps the creative in charge, keeps the science available and treats taste as a job worth protecting. Being able to point at those things on a trade-show floor felt rather good.

Ari sitting during IBC week

Give the creative a system that fits

I've been calling it Imaginary Tools. The longer description is creative computer clusters.

Start with the work you recognise. A brief. A shot list. The image you are trying to make. What must be true at the end, and what must never ship. Then arrange the compute around that.

AI can accelerate the labour. It does not need to become the author. If you've spent years learning how an image should feel, the useful tool is the one that lets you apply that judgement more clearly.

Four things keep me oriented: brief, taste, sensor truth, handoff. Capture, grade and delivery have to stay connected. Otherwise you spend the day reconstructing intent between windows.

Write the brief into a file. Bring the result back as a file. Leave the conversation available for the next decision. Files can be wrong too; their advantage is that you can inspect the actual contract instead of trying to remember which version of the chat mattered.

Taste stays with the person whose name is on the work. The system can prepare a release. The person decides whether it goes out.

Start with what must survive

Trade shows are excellent at giving an old job a new box. I still start from the sensor, the file and the intended image.

What has to survive this pass? Which information will the next person need? Where can the look change, and where have we committed it?

Those questions decide whether a tool belongs. The new box does not get to set the brief just because it looks lovely on a stand.

Twelve years on sets taught me those priorities. Building the tools in 2026 gave me a way to encode more of them. I arrived at IBC carrying both.

A product page, between talks

While I was on the IBC stage floor, we shipped a full redesign of the Pushing Squares product page.

I was still in the noise of it: Nikon talk, friends in the aisle, pointing at the screen like an excited gremlin. The page had to feel like us. That was the brief I held while the jobs did the work.

I talk to one Grok Bot. That is my seat. Around it sits a graph of jobs: jobs are nodes, handoffs are edges. A specialist gets a written brief when its kind of work is ready and returns a file for the next step.

That keeps me out of a row of competing conversations. The seat holds the brief. The jobs carry the labour. I keep the judgement and the publication decision. Nothing in this setup gets to publish itself.

The account uses the same arrangement described in Grok Bots made me stop prompting. Here it had a very concrete test: a product page we were willing to put in front of people that day.

We were mainly showing Subtrackt_ and CBA_: film-like colour subtraction and cinema-lens fringe. The marks spun in ASCII on the page, sitting with the image rather than turning the whole thing into a catalogue.

The six marks below are Crush_, Fuzz_, CBA_, Subtrackt_, Refrakt_ and Pause_. Have a fidget. They are part of that page's visual language; the toy does not run the plugins.

What a year of raw made me want

I had spent the past year shooting with the ZR. The flexibility of R3D raw kept surprising me: so much sensor information, and room for the image to move before the look became a fixed delivery.

That is the experience behind this part of the design. When there is still useful information in the negative, I want the handoff to preserve the ability to work with it. I do not want a finished-looking file to imply that every creative decision is finished too.

The file is not the look. It is where the look can still change.

It also needs a reliable home. A colour decision buried in a chat or an unexplained bake is a poor handoff. You may have the picture and still be missing what the next person needs to understand it.

That year on the ZR is the sensor side of .neg. A render engine is the other route into the design: a negative of its own, with the information the later work needs. The names below give that intention a shape. They are not evidence that a new format already interoperates with every tool.

The cine-specific deep dives belong at TGSD.dev: scopes, sensor data, what the raw does when you push it. This page stays with the journal. ❤️

Ari standing during IBC week

Name the handoff before it becomes folklore

Someone grades. Someone shares. The next person asks which science or alterations travelled with the picture. If the answer lives only in someone's memory, the file has left too much work behind.

Sometimes you need to keep the science. Sometimes you deliberately need a flattened delivery. Sometimes the recipient already has the negatives and only needs your alterations. Those are different jobs. Giving them different names makes the conversation shorter and more exact.

That is what I was doing with Candy at IBC. Naming the family so the build could be discussed clearly. This was a design account, not a shipped-format announcement.

Candy: five files, three modes

.neg is the digital negative in this design. Its own file, with a sensor or render-engine origin. It is not a sidecar or a metadata note about another negative.

.candy is the working file plus the science. It is where the work remains open. LOCK, TASTE and MIX are modes inside it, not separate .lock, .taste or .mix files.

.master is aware of delivery requirements and keeps the science. It is not flattened. The design lets it exchange with .candy in both directions, because a request for delivery should not silently erase the working relationship.

.share is the deliberate flatten for a foreign handoff. That path is one-way. Do not describe the easy-to-share result as if it can reconstruct the full working file.

.wrapper carries alterations for a peer who already has the .neg files. It comes off .candy; it is not another negative. Without the recipient's negatives, the alterations have nowhere to land.

The central relationship is:

.neg → .candy ↔ .master → .share

The separate wrapper path assumes shared negatives. Keep that requirement in the sentence whenever you explain it.

Inside .candy, the three modes separate intent from viewing conditions:

  • LOCK holds the intended colour or look.
  • TASTE represents this observer and this display: how the image is mapped for the person viewing it here.
  • MIX is the slider between intent and this observer/display, without pretending they are the same thing.

Those are the architectural intentions. Preserving colour through real tools and displays needs implementation and checks; a filename cannot do that work on its own.

One seat, work that can travel

I used to spend too much thinking inside the chat: ask, wait, ask again. The labour and the judgement fought for the same space.

The arrangement I am describing gives them somewhere to go. A brief in a file. A job with a clear result. A handoff another job can use. A check that gets fresh eyes. A decision that comes back to me.

That is how the product-page work could happen while I was still on the IBC floor. It is also the connection between Imaginary Tools and the Candy design: let the work travel without losing the person's intent on the way.

The last click is human. It is not a rota position for whichever model happens to be available.

Build the tool your industry has wished they'd had for decades.

That is what I am doing. Subtrackt_ and CBA_ spinning on the page. Friends laughing in the aisle. Nikon still in my throat a bit. A journal full of the next things to make.

I'll keep shooting, and I'll keep testing the containers against the work they need to carry.

//A<3

[ comments ]

guests welcome. members show first in the list

no comments yet. start the thread_

1 / 10