Introduction
Somewhere between a construction site in Rogaland, Norway, and a code editor lit up with TypeScript, React, and Python, a portfolio has been quietly taking shape. It belongs to a developer who goes by knoksen on GitHub — a builder in the most literal sense, with roughly three decades split between carpentry, building technology, and now, software. His repositories range from a dashboard for monitoring a dark-web language model to a ketamine-inspired light-show generator, a WordPress security toolkit, an AI-driven music streaming tool, and a research platform exploring what it might mean to build a “digital human.”
This article exists for a simple reason: portfolios like this one are becoming one of the most important artifacts of the AI era, yet they are rarely examined seriously. We are used to treating a GitHub profile as a résumé line — a link tucked at the bottom of a CV, glanced at for thirty seconds. But a repository list, read carefully, is closer to a field journal. It records what a person was curious about, what tools they reached for, what problems annoyed them enough to fix, and how their thinking changed over time. In knoksen’s case, that journal happens to belong to someone who spent fifteen years as a carpenter and fifteen more as a building technician before turning, seemingly with just as much seriousness, toward code.
The purpose of this piece is threefold. First, to trace how a portfolio like this comes into being — the historical and personal context that makes a “hobbyist” repository list meaningful rather than incidental. Second, to establish why this kind of portfolio matters right now, at a moment when the software industry itself is being reorganized by artificial intelligence, low-code tooling, and an explosion of first-time contributors. Third, to walk through the repositories themselves as case studies, extracting what each one teaches about cross-disciplinary development, and to look ahead at where a portfolio built this way might be headed next.
For readers who build things — whether in wood, concrete, code, or all three — this is also an argument: that the walls between trades and disciplines are far more permeable than they appear, and that the tools now available make crossing them easier than at any point in computing history.
Historical Context: How a Builder Becomes a Developer
The Slow Convergence of Trades and Technology
To understand a portfolio like knoksen’s, it helps to understand the profession it grew out of. Building technicians and construction professionals have always been quietly technical people. Reading blueprints, calculating loads, sequencing a build, and troubleshooting a failure under time pressure are all forms of structured problem-solving — the same cognitive muscles that show up in debugging a script or designing a database schema. What changed, over the last two decades, is not the underlying skill but the distance between the trade and the tools of software development. That distance has been collapsing.
In the 2000s, moving from a construction trade into software typically required years of formal retraining: a computer science degree, a bootcamp, or at minimum thousands of hours of self-study using dense textbooks and unfriendly tooling. Version control was arcane, package managers were fragmented, and the feedback loop between “I have an idea” and “I have something running” could stretch into weeks. A building technician with a good idea for, say, a site-monitoring dashboard had almost no realistic path to build it themselves.
The 2010s changed the economics somewhat, with the rise of accessible frameworks, Stack Overflow as a de facto textbook, and cloud hosting that removed the need to run a personal server. But the real inflection point — the one that explains a portfolio spanning language-model dashboards, IoT-adjacent camera systems, and generative art tools, all built by one part-time hobbyist — is the AI-assisted coding wave that began in earnest around 2021 with the first code-completion tools and accelerated sharply from 2023 onward.
GitHub Copilot and the Democratization of “Who Can Build Software”
GitHub’s own research gives a useful anchor for this shift. The company’s Octoverse 2025 report describes 2025 as GitHub’s fastest absolute growth year in its history, with more than 36 million new developers joining the platform — over one every second, on average — bringing the total past 180 million (GitHub, 2025). GitHub explicitly attributes part of this acceleration to the release of GitHub Copilot Free in late 2024, which coincided with a sharp step-change in sign-ups beyond what prior growth trends had predicted (GitHub, 2025). March 2025 alone brought 255,000 first-time open-source contributors, the largest single month in the platform’s history (GitHub, 2025).
This matters for a portfolio like knoksen’s because it is precisely the kind of profile that this democratization was expected to produce: a domain expert from a non-software field, equipped with AI-assisted tooling, building working software across a surprising range of domains without the traditional years-long apprenticeship. Where a repository list from 2015 might have shown a narrow, deep specialization (the natural output of years of formal training), a 2025-era hobbyist portfolio can plausibly span web dashboards, desktop applications, generative art, and light experimental research — because the barrier to trying is so much lower than it used to be.
The Personal Timeline Behind the Repositories
Within that broader industry shift, knoksen’s own trajectory follows a recognizable arc. Profile information associated with the account describes its owner as a “Building Technician — unofficial Computer Engineer,” educated at a master’s level in building construction, with computing pursued “mostly hobby based” alongside “some certificates.” He is linked to two Norway-based organizations, Terratek-AS and Taksdal-BT, both oriented toward construction and technical building services, and to a personal umbrella site, jarlhalla.no, that gathers his various ventures. The GitHub account itself sits alongside presences on ORCID, Pinterest, and DeviantArt — a spread that signals someone treating software, research identity, and visual art as parts of the same creative practice rather than separate compartments.
That is the historical backdrop worth holding in mind while walking through the repositories themselves: this is not a computer-science graduate’s portfolio, curated to demonstrate algorithmic fluency to a hiring committee. It is closer to a workshop bench — the accumulated tools of someone who builds things because building things is the point, and who happens to have picked up software as a new material somewhere along the way.
Current Relevance: Why This Kind of Portfolio Matters Now
The Numbers Behind the Moment
It is worth pausing on just how large the ecosystem around portfolios like this one has become. GitHub’s Octoverse 2025 data reports 395 million public repositories hosting 1.12 billion contributions and 518.7 million merged pull requests in the preceding year — each a record (GitHub, 2025). Six of the ten fastest-growing open-source projects by contributor count were AI infrastructure projects, and more than 1.1 million public repositories now incorporate a large-language-model software development kit, representing 178% year-over-year growth (Forbes, 2025). Monthly contributors to generative-AI projects rose from roughly 68,000 in January 2024 to about 200,000 by August 2025 (Forbes, 2025).
Put simply: a hobbyist portfolio that mixes a BERT-style dashboard, an AI-controlled music tool, and experimental “digital human” research is not an eccentric outlier anymore — it is a recognizable pattern of the current moment, sitting squarely inside the fastest-growing segment of the fastest-growing developer population in GitHub’s history.
The Rise of the Citizen Developer
A second, related trend gives this portfolio additional context: the rise of the “citizen developer” — someone building real software without formal training as a professional engineer. Analyst estimates vary in their specifics, but they agree on direction and scale. Industry trackers estimate that citizen developers now number in the tens of millions worldwide, with counts commonly cited around 16 million and growing at double-digit rates year over year (ToolJet, 2026; Searchlab, 2026). Gartner-derived figures repeatedly put non-professional builders at roughly four times the number of professional developers inside organizations that have formally embraced low-code and citizen-development programs (Kissflow, 2026; Integrate.io, 2026). Separate Gartner-sourced projections suggest that by the middle of this decade, a majority of new business applications will be built at least partly outside formal IT departments (Quixy, 2025).
Knoksen’s portfolio is, in effect, an independent, non-enterprise instance of exactly this trend: a domain professional from construction, building a working software portfolio without the institutional scaffolding — bootcamps, employer-sponsored training, IT department oversight — that usually accompanies this kind of career-adjacent skill growth. What is happening inside large companies under the banner of “citizen development” is also happening, informally and independently, on personal GitHub accounts around the world.
Challenges That Come With This Trend
None of this is friction-free. GitHub’s own Open Source Survey data, cited in recent reporting, finds that 93% of open-source users consider incomplete or outdated documentation a significant problem, while 60% of contributors rarely or never help maintain it (Rockstar Developer University, 2026). The same analysis notes that a majority of maintainers work unpaid, with burnout cited as a leading reason contributors consider stepping away (Rockstar Developer University, 2026, citing Tidelift’s Maintainer Survey). For a solo, hobby-driven portfolio like knoksen’s — spread across many small, independently maintained projects rather than one flagship product — these are exactly the pressures to watch: enthusiasm for starting new repositories is abundant, but the unglamorous work of documentation, security patching, and long-term maintenance is where most personal open-source portfolios quietly stall.
Practical Applications: The Repositories as Case Studies
The most useful way to read this portfolio is not as a list, but as a set of case studies, each illustrating a different way that a building-trades background and an AI-assisted development style combine in practice.
Case Study 1 — BERT-Dashboard / darkbert-dashboard: Interfaces for Interpreting AI Models
One of the more technically ambitious repositories in the portfolio is a web-based dashboard, built with TypeScript and React, described as a tool “for visualizing and monitoring DarkBERT model performance and outputs.” DarkBERT itself is a real, peer-reviewed research artifact: a transformer-based language model pretrained specifically on Dark Web text, developed by researchers at KAIST and the cybersecurity firm S2W and published at the 2023 Annual Meeting of the Association for Computational Linguistics (Jin et al., 2023). The original paper demonstrates that a domain-specific language model trained on Dark Web data outperforms general-purpose models on tasks like classifying underground forum activity and detecting ransomware or data-leak sites (Jin et al., 2023).
Building a dashboard around a model like this is a genuinely useful, non-trivial application. Specialized security and threat-intelligence models are notoriously hard for non-specialists to interrogate — their outputs are numeric, their behavior opaque, and their practical use requires visualization layers that translate model confidence scores into something a human analyst can act on. A dashboard project of this kind sits at the intersection of front-end engineering (React component design, TypeScript type safety) and applied machine learning literacy (understanding what a classification score from a domain-specific BERT variant actually represents). For a developer whose formal training is in building construction rather than computer science, choosing to build tooling around a cybersecurity research model — rather than simply consuming an existing dashboard — is a meaningful signal of technical ambition that extends beyond routine web development.
Case Study 2 — knokspack: Practical Infrastructure for the Ordinary Web
Not every project in the portfolio aims at the research frontier. Knokspack is described as an “all-in-one toolkit to secure, optimize, and grow” WordPress websites — the kind of unglamorous, high-utility tooling that keeps millions of ordinary small-business websites online. WordPress remains one of the most widely deployed content-management systems in the world, and tooling that bundles security hardening, performance optimization, and growth features addresses a real and continuous need: most WordPress site owners are not developers, and configuring security plugins, caching layers, and SEO settings individually is a recurring source of friction.
This project is worth highlighting precisely because it is the least exotic item in the portfolio. It demonstrates a builder’s instinct: rather than only chasing the most technically glamorous problem (a dark-web model dashboard, a light-show generator), the same portfolio also contains a project aimed squarely at solving a mundane, recurring, practical problem for non-technical end users — the same instinct that, in a construction context, shows up as fixing the boring-but-critical parts of a building (moisture barriers, load paths) rather than only the visible finishes.
Case Study 3 — knoksDJ: AI-Directed Audio Streaming
Described as a “music streaming system with text prompt control capabilities,” built primarily in Node.js, knoksDJ sits at the intersection of two rapidly evolving fields: real-time audio streaming infrastructure and natural-language-driven control interfaces. Rather than a conventional media player controlled by clicks and sliders, a “text prompt control” system implies an interface where a user can describe, in natural language, what they want to happen — a mood, a genre shift, a specific instruction — and have the system interpret and execute it.
This is a small-scale but telling example of a much larger industry pattern: the move away from rigid, menu-driven software controls toward conversational and generative interfaces, mirroring how large language models have reshaped user expectations across almost every category of consumer software since 2023. Building this kind of interface, even at hobbyist scale, requires stitching together audio streaming protocols, a Node.js backend, and some form of natural-language interpretation layer — a genuinely cross-disciplinary integration exercise.
Case Study 4 — earthnullschool: Learning by Forking
Among the portfolio’s repositories is a fork of the well-known “earth” project, originally created by developer Cameron Beccario, which visualizes global weather conditions by rendering wind, ocean currents, and atmospheric data as flowing animated maps. The original project’s own documentation is candid about its purpose: it describes itself as “a personal project I’ve used to learn JavaScript and browser programming,” built on the earlier “Tokyo Wind Map” project, and explicitly invites feedback and contributions “especially those that clarify accepted best practices.”
Forking an established, well-documented open-source project for the explicit purpose of learning is one of the oldest and most effective on-ramps into software development, and its presence in this portfolio is instructive. Rather than every project being an original concept, part of the portfolio’s value lies in the deliberate choice to study a well-built system from the inside — reading its data pipeline, its rendering logic, and its server architecture — which is a direct software analogue to a building technician studying an existing structure’s blueprints before attempting an original design.
Case Study 5 — KH0L3: Generative Visual Art as a Software Problem
KH0L3 is described as a ketamine-inspired light-show application, positioned within the portfolio’s broader creative and digital-art strand alongside the developer’s Pinterest and DeviantArt presences. Generative light and visual art tools are a well-established genre in creative coding, typically combining real-time graphics rendering (shaders, particle systems, or canvas-based animation) with parameters that produce psychedelic, flowing, or reactive visuals — often synchronized to audio input, similar in spirit to decades of “visualizer” software built for music players and live performance visuals.
What makes this entry notable in context is less the specific creative theme and more what it demonstrates structurally: that the same GitHub account houses both a serious applied-AI dashboard project (BERT-Dashboard) and an explicitly artistic, mood- and sensation-driven creative-coding project. That combination reflects a working style in which software is treated as a general-purpose creative and technical medium — the same posture a building technician might take toward physical materials, using the same underlying skills (structural thinking, iterative testing, tolerance for failure) toward both functional and expressive ends.
Case Study 6 — RoomZero / ReDigitalBeing: The Portfolio’s Most Ambitious Undertaking
The most architecturally serious project associated with this developer is RoomZero, subtitled the “Digital Human Research Environment,” built around an internal research agent named Eir. Its technical stack — FastAPI for the backend, SQLite for storage, a progressive web app front end, a command-line chat interface, a packaged Windows installer, GitHub Pages for public documentation, and Unreal Engine integration bridged over WebSockets — represents a substantially more complex undertaking than any single-purpose dashboard or utility elsewhere in the portfolio. The project has reached a tagged 1.0.1 release, expanded its automated test suite from 16 to 47 passing tests, and maintains structured weekly reports covering platform status, the Unreal Engine–WebSocket integration, simulation events, and near-term milestones.
This project is the clearest evidence in the portfolio of a shift from “hobby experiments” toward something closer to a sustained research program. Multi-service architecture, real-time bridging between a Python backend and a 3D game engine, a persistent test suite, and a formal reporting cadence are all hallmarks of engineering discipline that go well beyond a weekend project. It also reflects one of the more philosophically ambitious currents in current AI research — attempts to give language-model-driven agents a persistent “body” or environment, rather than treating them as stateless chat responders — echoing broader industry experimentation with embodied and simulated AI agents throughout 2024 and 2025.
What the Case Studies Add Up To
Read together, these six projects sketch a coherent creative identity rather than a scattershot list. There is applied AI tooling (BERT-Dashboard), practical infrastructure for ordinary users (knokspack), conversational/generative interface design (knoksDJ), disciplined learning-by-forking (earthnullschool), expressive creative coding (KH0L3), and a genuinely ambitious systems-architecture research project (RoomZero). Few professional software teams would deliberately spread their public output across such varied terrain — but a solo, curiosity-driven builder, unconstrained by a product roadmap or an employer’s domain focus, naturally produces exactly this kind of range.
Future Implications: Where Portfolios Like This Are Headed
AI-Assisted Solo Development Will Keep Expanding the Realistic Scope of One Person’s Portfolio
The trajectory of GitHub’s own data suggests this pattern will intensify rather than fade. With TypeScript overtaking Python and JavaScript as the platform’s most-used language in 2025, driven substantially by AI tooling’s affinity for typed codebases (GitHub, 2025), and with over a million public repositories now integrating large-language-model SDKs (Forbes, 2025), the tooling available to a solo, cross-disciplinary developer is becoming both more powerful and more standardized. A portfolio like this one — spanning a research-grade dashboard, a generative art tool, and a multi-service AI research platform — was difficult to imagine a single hobbyist producing a decade ago. The current trajectory of AI-assisted coding tools makes it increasingly plausible that the range of such a portfolio, not just its depth in any one project, will keep expanding.
The Low-Code and Citizen-Development Wave Will Blur the Line Between “Trade” and “Software” Further
Market analysts project the global low-code/no-code market to be worth somewhere between roughly $44 billion and $65 billion in 2026, depending on methodology, with growth rates commonly cited between 19% and 26% annually (Caspio, 2026; ToolJet, 2026). Some projections put the citizen-developer population as high as 16 million worldwide in 2026, up sharply year over year (ToolJet, 2026; Searchlab, 2026). If that trajectory holds, the boundary that historically separated a “trade professional who occasionally uses software” from a “software developer” will likely continue to dissolve — not only inside enterprises adopting formal citizen-development programs, but for independent builders maintaining personal portfolios exactly like this one, who increasingly treat coding as a natural extension of a technical trade background rather than a separate career.
Research-Grade Ambitions, Like RoomZero, Point Toward a Coming Wave of Independent AI Research
Perhaps the most interesting forward-looking signal in this portfolio is the RoomZero project’s trajectory from hobby experiment toward disciplined research platform, complete with a formal testing regime and structured status reporting. As foundational AI tooling becomes more accessible and better documented, it is increasingly plausible for independent researchers and hobbyists — without institutional lab affiliations — to pursue genuinely open research questions, such as how to give an AI agent a persistent, embodied presence inside a simulated environment. Whether or not any specific project like RoomZero reaches a wide audience, the pattern it represents — independent, well-documented, test-covered AI research conducted outside traditional academic or corporate structures — is likely to become more common, not less, as the tooling for rigorous solo research continues to mature.
The Maintenance Challenge Will Not Disappear on Its Own
At the same time, the structural challenges identified earlier in this analysis — thin documentation, unpaid maintenance labor, and burnout risk (Rockstar Developer University, 2026) — apply just as much to ambitious solo portfolios as to large collaborative projects, arguably more so, since a solo maintainer has no team to distribute the load. A portfolio spanning six or more actively maintained repositories, several with real external dependencies (language models, game engines, streaming protocols), faces a genuine long-term sustainability question: which projects get sustained attention, and which quietly stall once the initial creative spark fades. How individual builders like the one behind this portfolio choose to prioritize — consolidating effort into a flagship project like RoomZero versus continuing to spread output across many smaller experiments — will likely determine which parts of the portfolio have lasting impact versus which remain interesting but ultimately abandoned artifacts.
Conclusion
A GitHub repository list is easy to skim past, but read closely, knoksen’s portfolio tells a coherent and genuinely current story. It begins with three decades in construction and building technology, a field built on structured problem-solving that translates more directly into software development than is usually assumed. It unfolds against the backdrop of the fastest developer growth in GitHub’s history, driven substantially by AI-assisted coding tools that have made it realistic for a domain expert from an unrelated trade to build a dashboard for a cybersecurity research model, a WordPress security toolkit, a conversational music tool, a learning exercise built on a respected open-source weather visualization, an expressive generative art application, and — most ambitiously — a persistent, tested, actively documented AI research platform.
Each of these projects, examined individually, is a modest case study in applied engineering. Examined together, they describe something larger: an early, personal instance of a trend that is reshaping the software industry at scale — the rise of the cross-disciplinary, AI-assisted, largely self-taught builder, operating outside the traditional boundaries of formal computer-science training. The open questions that remain are not really about talent or ambition, both of which this portfolio demonstrates clearly, but about sustainability: which of these threads will receive the sustained maintenance needed to mature into lasting tools, and which will remain valuable primarily as evidence of a genuinely versatile, curious way of building. Either way, portfolios like this one are worth studying seriously — not as résumé filler, but as a real-time record of how the walls between trades and technology are coming down.
References
Forbes. (2025, November 1). 10 key takeaways from GitHub Octoverse 2025 report. https://www.forbes.com/sites/janakirammsv/2025/11/01/10-key-takeaways-from-github-octoverse-2025-report/
GitHub. (2025). Octoverse: A new developer joins GitHub every second as AI leads TypeScript to #1. The GitHub Blog. https://github.blog/news-insights/octoverse/octoverse-a-new-developer-joins-github-every-second-as-ai-leads-typescript-to-1/
Jin, Y., Jang, E., Cui, J., Chung, J.-W., Lee, Y., & Shin, S. (2023). DarkBERT: A language model for the dark side of the internet. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) (pp. 7515–7533). Association for Computational Linguistics. https://aclanthology.org/2023.acl-long.415/
Kissflow. (2026). 65+ no-code statistics 2026: Market size, growth & adoption data. https://kissflow.com/no-code/no-code-statistics-2026/
Integrate.io. (2026). No-code transformations usage trends — 45 statistics every business leader should know in 2026. https://www.integrate.io/blog/no-code-transformations-usage-trends/
Quixy. (2025). Game-changing top 60 no-code low-code citizen development statistics. https://quixy.com/blog/no-code-low-code-citizen-development-statistics-facts/
Rockstar Developer University. (2026). Open source contribution statistics 2026: Contributor data. https://rockstardeveloperuniversity.com/open-source-contribution-statistics/
Searchlab. (2026). No-code & low-code statistics 2026: 50+ data points & insights. https://searchlab.nl/en/statistics/no-code-low-code-statistics-2026
ToolJet. (2026). Low-code statistics 2026: 60+ facts, figures & trends business leaders need to know. https://blog.tooljet.com/low-code-statistics-market-ai-trends-2026/
Caspio. (2026). The state of no-code in 2026: Market trends and what’s next. https://www.caspio.com/blog/state-of-no-code-2026/
GitHub profile and repository descriptions referenced throughout: github.com/knoksen, including BERT-Dashboard/darkbert-dashboard, knokspack, knoksDJ, earthnullschool (forked from cambecc/earth), KH0L3, and RoomZero/ReDigitalBeing.