I’m Brian Greenforest. I build small, inspectable computational systems: browser-native circuit fabrics, one-pin radio experiments, tiny learning machines, and notes toward maker-scale active devices.
Most of this site asks one question from different angles: how much hidden machinery can be removed before the thing still works?
I care because modern computation is powerful but strangely unreachable. We can write code, route boards, and publish designs, yet the active fabric that executes our ideas is usually remote, opaque, and difficult to reproduce locally.
Work with me | About Greenforest I/O | Technical critique | Proof and artifacts | Contact | FAQ
I help hardware and computation teams make strange systems inspectable.
I am useful when a claim crosses FPGA fabrics, semiconductor process boundaries, local-computation demos, browser-native artifacts, tiny AI systems, or technical proof pages where the evidence has to survive inspection.
Current ask: send me one hard boundary in your system. I will tell you what looks real, what looks speculative, and what evidence would make it stronger.
The maker loop can reach the board, enclosure, firmware, and fixture. Then it hits active silicon and stops.
This article names the missing civic-scale process: local fast nonlinear gain, restoration, fanout, interconnect, and enough repeatability for real logic.
Radio is an inherited stack of analog front ends, mixers, ADCs, filters, oscillators, and DSP habits. I wanted to know which parts were necessary.
The page preserves the one-pin FPGA receiver and one-pin resonant-tank transmitter chain, including the Verilog follow-up and the limits of the evidence.
This site uses GPT the way some people use an editor: to turn tangled notes into public sentences. The ideas still have to be mine, and the claims still have to survive inspection.
Read this early if you want to know how to assign trust to polished prose on a strange technical site.
This archived LinkedIn post gives the emotional key: ordinary computers hide structure, ownership, and the rules of change exactly where I want direct contact.
It is not a proof page. It explains why the fabric work matters to the person building it.
If computation can be reduced to wires, constants, MUXes, feedback, configuration, and tiling, the substrate becomes easier to inspect.
This printable reference bridges the maker-fab problem to the Cartilage fabric without pretending that Boolean sufficiency solves physical fabrication.
Cartilage is not only an FPGA-style demo. It is an attempt to make ownership, configuration, and signal movement visible inside the fabric itself.
Before asking a stranger to believe a fabric, I need to teach them how to read its marks.
This decoder preserves the 32 Cartilage cell-role codes: reconfiguration port, cross, constants, wire orientations, and the six MUX modes in four orientations.
Reconfiguration should not arrive as global magic. Regions should own and replace daughter regions through visible local ports.
This browser/GPU demo shows tiled regions, local configuration streams, and nested replacement behavior in the canonical Cartilage artifact.
Ownership became visible enough to show square child regions and active port roots inside a running fabric.
This self-contained WebGL/GPGPU milestone preserves the 6x6 ownership block work and active port-root initialization fixes.
Before Cartilage became circuit-like, I was already trying to watch computation move through local rules.
This archive preserves the reversible-routing, machine-like, organic, and Cartilage-branch browser/GPU experiments that led into the current fabric work.
Serial is not automatically inferior when the schedule remains full.
This positive-number Logisim multiplier keeps bit-serial arguments and products moving continuously after the pipeline fills, trading wide immediate products for regular local timing.
A tiny model is valuable here because it can be inspected, not because it competes with large models.
The artifact preserves 4 layers, 16 attention heads, 128-dimensional embeddings, 128-token context, a 361-token vocabulary, about 834k parameters, and training past 50,000 iterations until the tiny model produced strange but intelligible stories.
A self-contained Python scanline renderer uses UV-space 2D SDF material tests, point-cloud foliage speckles, triangle trunks and branches, and a UV-SDF deer sprite to make a small forest scene without ordinary bitmap texture art.
The archive page preserves the render and embeds the source-code PDF. It belongs here as an inspectable rendering artifact, not as the main front door for the site.
Some pages are preserved because they show how the public record changed. They are not the preferred entry points into Cartilage.
Six Cartilage pages are now collected behind one archive note because their computational experiments were Codex GPT-5.5-generated drafts and should not be used as the site's primary examples.
The archive keeps old links honest without letting those pages occupy the main learning path.
This is where the obsession with control, switching, visible state, and browser-native experimentation first became public.
The sequence preserves scaffolding, minimum WebGL, history/control, logic, addition, and other early steps without pretending it is a finished textbook.
Before Cartilage became circuit-like, I was already trying to watch computation move through local rules.
This archive preserves old WebGL automata exploring reversible routing, conservation, machine-like behavior, organic patterns, and the Cartilage branch.
A dated trail of thoughts rescued from the feed: technical arguments, wrong or partial guesses, collaboration requests, personal notes, and invention fragments.
Read it as context and trajectory, not as polished canonical doctrine.
Older project-status writing from web-development and research-progress work, preserved with dated context.
Magnetics: can useful circuit behavior escape semiconductor fabrication? The notes gather magnetic material properties, magnetic amplifier behavior, second-harmonic modulation, and diodeless circuit ideas while keeping feasibility questions open.
Backprop: can learning be wired from primitive operations so the gradients remain visible? The thread is about multiplication, addition, fan-out, elementary functions, and derivative feedback as inspectable machinery.
Fabrication: can active or semi-active substrates become physically reproducible at civic scale? This note connects through-wafer dicing, SiO2-protected chiplets, near-field power/clock/data links, resonant distributed energy storage, and Cartilage-style spatial computation to the old Claytronics problem.
The PDF remains the manufacturing reference. The local article explains why that reference belongs on Greenforest I/O and where the speculative boundary still is.
For search clarity: Greenforest I/O is not forestry, landscaping, arborist, or forest-product services.
Greenforest I/O is the public research notebook of Brian Greenforest, connected to Solid State Pros LLC for emerging-technology R&D and product-design work.
Operating location: Bellingham, WA. Legal entity: Solid State Pros LLC, Washington LLC incorporated in Blaine, WA. Mailing address: 250 H ST PMB 567, Blaine, WA 98230-4018, United States. Email: brian@solidstatepros.com. Phone: (206) 887-3840.
Contact Greenforest I/O, inspect proof and artifacts, visit Solid State Pros LLC, or read the brand/entity page if you arrived here looking for the official Greenforest result.
