Casting a chip is the moment a working circuit stops being a one-off. Instead of rebuilding the same gates for the next contract, you cast the design once, keep it in a catalogue, and drop the finished part into everything you build afterwards.
After the NAND gate, this is the second skill that decides your pace through IT THINKS!. The campaign spans 62 contracts across six eras, and the players who finish fastest are not the ones who wire the cleanest circuits — they are the ones who stop rewiring.
What casting actually does in IT THINKS!
The Steam page describes it in one line: "cast designs into reusable chips" and "build a catalogue of your own components." That is the whole idea. A circuit you have already proven on the test bench gets frozen into a chip, and the chip behaves like any other part: one block, defined inputs, defined outputs.
The point is not saving parts. The point is saving attention. A half adder is a dozen-plus gates, and once it is a chip you no longer have to think about it while you are designing the thing that uses it. That is how a machine that fills a board stays a design you can still reason about.
When to cast, and when to keep building
Three situations where casting pays off immediately:
- You have built the same thing twice. The second build is the signal. Anything you wire once for arithmetic will be needed again in memory, in the processor, and in the matrix pipeline.
- The circuit is proven, not just plausible. Cast after the test bench passes, not before. A broken chip gets copied into every later design and you will find it six contracts later.
- You are about to run out of room to think. If the board is so dense that you cannot trace an output by eye, cast the sub-blocks you already trust and the picture gets simple again.
Keep building by hand when the circuit is one you are still tuning. A chip freezes the design; if you are about to trade area for speed, or speed for power, you want that decision made before you cast, not after.
The trade-offs casting cannot hide
IT THINKS! asks you to balance area, speed, memory bandwidth, power, and parallel efficiency. A chip makes those trade-offs permanent for that component:
- Area. A chip still occupies board space, and a cast design that wastes gates wastes them everywhere it is used.
- Speed. Signal depth does not disappear when you cast. A technically correct chip with ten gates of depth is still slow, and later contracts start caring.
- Power. Once chips are reusable, you tend to use more of them. Power is the bill for that habit.
This is also why licensed parts and cast parts are different tools. A license gets you a working part now. A cast chip gets you a part that is yours and that you can revise — and revising your own chip, instead of rebuilding every circuit that used it, is the whole return on the habit.
How to organise a component catalogue
A catalogue becomes useless around the time it becomes large, and it gets large fast. Three habits keep it readable:
- One job per chip. A chip that both adds and stores is a chip you will hesitate to reuse.
- Name by what it does, not when you made it. "4-bit adder" is a part you will find again in era five. "Adder v2" is not.
- Cast the boring parts first. Inverters, AND gates, OR gates, and the half adder carry the most reuse per unit of effort. Exotic logic gets used once and then sits in the list.
The mistakes that cost the most
- Casting an unverified circuit. The most expensive error in the game, because every later design inherits the bug.
- Casting too early in a design you still plan to change. You either live with the worse version or rebuild everything that used it.
- Never casting at all. You will still finish, but you will spend the middle of the campaign rewiring gates instead of designing a machine.
- Casting what you could license for free. If a contract only ever needs the part once, the factory version is the cheaper decision.
From chips to a machine that thinks
The eras are a chain of ever larger chips: arithmetic, then memory, then a programmable processor, then parallel cores, then matrix multiplication, then a chip that recognises the shapes you draw. Each era's work becomes the next era's parts list.
That is why casting is worth learning properly on your first pass. Everything after arithmetic is a composition problem, and composition only works when the lower layer is something you trust enough to stop looking at.
If the logic-gate layer is still shaky, start with the NAND gate truth table — every chip in your catalogue sits on top of it. If you are stuck on a particular contract instead, the contract walkthrough goes era by era.
When you want to experiment without a client waiting, the game's free bench has buttons, a keypad, seven-segment displays, a 16×16 light array, a buzzer and dice, with a front panel and room for up to 4,000 parts — and builds you make there can be shared as links anyone can open and press keys on in a browser, no account needed. You can also try the demo, which is free and needs no download.
FAQ
Do I have to build every part from NAND in IT THINKS!?
No. Every part can be built from NAND or licensed from the factory, and you can come back and build your own version later. Licensing is how the game lets you keep momentum when a contract does not need you to learn that part yet.
Does casting a chip make it faster?
No. Casting freezes a design; it does not improve it. If a component is slow because of signal depth, the chip will be slow every time you use it. Fix the design before casting.
Can I revise a chip after casting it?
Yes — your catalogue is yours, and a cast part is meant to be revisited. The reason to cast carefully is that revisions cost you a sweep through the designs that already use it.
What is the free bench for?
Sandbox building with no contract deadline: the store page lists buttons, a keypad, seven-segment displays, a 16×16 light array, a buzzer, dice, and a front panel, with support for up to 4,000 parts and shareable links.