Spectel: Reviving Teletext for Ham Radio in the Age of Vibe Coding
Published: April 22, 2026
By Senior Technical/Financial Audit Journalist
Introduction: The Resurrection of a Dead Broadcast Standard
On the 40th anniversary of Aertel's test transmissions—the Irish national teletext service that once delivered news, sports scores, and television schedules to millions—a single developer has resurrected the underlying broadcast standard for an entirely different medium: amateur radio. The project, called Spectel, transmits teletext-style 40x24 character grids over VHF/UHF and HF bands using the AX.25 protocol, a feat accomplished not by a team of embedded systems engineers but by one individual leveraging Anthropic's Claude through "vibe coding" (Source 1: Primary Data - Project Developer Documentation).
The core question demands examination: why deploy a 40-year-old character grid over modern radio links when digital modes such as FT8, JS8Call, and DMR offer superior error correction, higher throughput, and mature ecosystem support? The answer reveals a hidden economic logic that extends well beyond amateur radio. Vibe coding—AI-assisted software development where the developer describes intent rather than writing syntax—has collapsed the cost of resurrecting legacy protocols to near zero, making niche technical experiments viable where they would otherwise remain economically irrational.
---
Section 1: Teletext 101 – The 40x24 Grid That Refused to Die
Teletext, first developed in the United Kingdom and rolled out by the BBC under the name Ceefax, was not merely a digital information overlay on analog television—it was an engineering compromise that became a cultural institution. Each teletext screen uses a 40-by-24-character grid, with one kilobyte of memory storing a full page of multicolor text (Source 2: Technical Specification - BBC Engineering Division). This was a deliberate constraint: broadcasters needed to transmit usable information within the vertical blanking interval of analog television signals, where spare scan lines carried approximately 1.5 kilobits per second.
The SAA5050 character-generator chip, designed by Mullard (a Philips subsidiary), stored characters in 6-by-10-pixel cells and interpolated them to an effective 10-by-18-pixel resolution for alphanumeric display. For graphic characters—the blocky, low-resolution images that defined teletext's visual identity—the chip used a 2-by-3 set of blocky pixels known as sixels (Source 3: Hardware Architecture Documentation - Semiconductor Data Sheets). This was a hardware hack that defined an era: the chip was cheap, deterministic in timing, and required no frame buffer.
The cultural anchor of this technology was the BBC Microcomputer, which carried the SAA5050 chip on its motherboard. This allowed home programmers in the 1980s to hijack teletext hardware for interactive applications, including educational games such as "Granny's Garden" and "The Oregon Trail" (Source 4: Retro Computing Archives - BBC Micro User Guide). The Commodore 64, by contrast, lacked native teletext hardware, necessitating software-based emulation that was slower and less reliable.
It is critical to note that teletext was designed for analog broadcast—unidirectional, broadcast-only, with no return channel and no error correction. Spectel's re-appropriation for two-way radio is a deliberate engineering choice, not a direct port. The developer has explicitly stated: "Inspired by how the BBC Micro co-opted teletext for its own purposes, I thought it might make a great radio protocol" (Source 1).
---
Section 2: The Radio Twist – AX.25, Baud Rates, and the 11-Second Screen
The AX.25 protocol, derived from the X.25 packet-switching standard, encodes ones and zeros as audible tones transmitted over radio frequencies. It is the backbone of amateur radio packet communication, used for everything from APRS position reporting to bulletin board systems. Spectel embeds teletext page data within AX.25 frames (Source 5: Protocol Specification - AX.25 Link Access Protocol for Amateur Packet Radio).
The transmission mathematics are straightforward. At 1200 baud—the standard rate for VHF/UHF packet radio—a single 40x24 teletext screen consumes approximately 960 bytes of payload data plus protocol headers. The transmission time is approximately 11 seconds. At 300 baud—the rate used for HF (high frequency) long-distance links—the same screen requires approximately 44 seconds (Source 1: Calculated from AX.25 Frame Overhead and Baud Rates).
Context matters: 11 seconds is comparable to the download speeds experienced by early dial-up teletext services in the 1980s, where users waited 10-15 seconds per page (Source 6: Historical Broadcast Engineering Reports - IBA Technical Review). This is not a regression; it is a deliberate trade-off between speed and robustness. VHF/UHF links at 1200 baud provide reasonable throughput for local communication within a 50-kilometer radius. HF links at 300 baud, while slower, ensure signal reliability over intercontinental distances where ionospheric propagation introduces multipath distortion and fading.
The developer's own verbatim assessment on HF performance: "Over VHF/UHF it works really well. Over HF... we can decode the signal to a certain extent, but the screen is taking 44 seconds to scan out" (Source 1). This suggests that Spectel's primary utility resides in local-area VHF/UHF networks, where the 11-second refresh rate is acceptable for static information display—weather bulletins, event schedules, or emergency coordination messages.
---
Section 3: The Vibe Coding Factor – When AI Makes Niche Projects Economically Rational
The most analytically significant dimension of this project is not the technology itself but the development methodology. The developer consumed approximately 250,000 to 300,000 tokens of Anthropic's Claude over several nights to produce the Spectel codebase (Source 1). At current API pricing for Claude (approximately $0.015 per 1,000 input tokens and $0.075 per 1,000 output tokens for Claude 3.5 Sonnet), this represents a total cost of roughly $15 to $35 for a working software-defined radio protocol implementation.
This cost structure is transformative. Previously, resurrecting a legacy broadcast standard like teletext for amateur radio would have required: (a) deep expertise in both analog video standards and radio-frequency digital protocols; (b) weeks of development time for protocol encoding/decoding, synchronization, and error handling; (c) potentially custom hardware for SAA5050 emulation. Vibe coding collapses these barriers. The developer explicitly stated: "Being honest with myself, I doubt this system—which I'm calling Spectel—would ever have come about without vibe coding" (Source 1).
The economic logic proceeds as follows: When development cost approaches zero, projects with near-zero addressable market become rational to pursue. The total global user base for ham radio teletext is unlikely to exceed a few thousand enthusiasts. A traditional development cost of $10,000-$50,000 would be irrational for such a niche. At $15-$35 in token costs, the project clears the economic threshold for "hobbyist viability."
However, the developer also expressed significant reservations: "I remain deeply ambivalent about vibe coding" (Source 1). This ambivalence is not sentimental but structural. Vibe coding produces code that the developer may not fully understand, creating a maintenance liability. If Claude's training data lacks specific knowledge of SAA5050 timing constraints or AX.25 frame sync edge cases, the resulting code may have latent bugs that only appear under specific RF propagation conditions—bugs that a human developer with deep domain knowledge might have preempted.
---
Section 4: Retro Nostalgia Meets Low-Bandwidth Necessity
The cultural dimension of Spectel cannot be dismissed as mere nostalgia. The developer's family accessed Aertel multiple times daily throughout the 1980s and 1990s in Ireland (Source 1). This is not an isolated anecdote; teletext services in the UK (Ceefax), Ireland (Aertel), and across Europe provided news, weather, stock prices, and travel information to households that lacked internet access. The services persisted into the 2010s before being discontinued as digital broadcasting rendered analog transmission infrastructure obsolete.
The emotional resonance of teletext's distinctive visual aesthetic—the blocky sixel graphics, the garish color palette, the fixed-width character grid—creates a basis for community adoption that purely utilitarian protocols lack. FT8 is optimized for weak-signal communication but is visually sterile. JS8Call adds conversational messaging but has no graphical component. Spectel offers a recognizable visual language that signals "this is teletext" to anyone who grew up with Ceefax or Aertel, reducing the learning curve for new users.
From a technical standpoint, the 40x24 grid also offers advantages for low-bandwidth communication that modern protocols have abandoned. At 1 KB per page, a single teletext screen can convey approximately 960 characters of information. For comparison, the typical FT8 transmission carries exactly 75 bits (13 characters) in 15 seconds. Spectel at 11 seconds per screen has a raw data rate of approximately 87 bytes per second—significantly more than FT8's 5 bytes per second, but with correspondingly higher signal-to-noise requirements.
---
Market Implications: The Future of Low-Bandwidth Digital Communications
The Spectel project, while niche, suggests three structural trends in technical communications development.
First, AI-assisted coding is democratizing protocol development for legacy and niche standards. The ability to generate working implementations of obscure protocols—teletext, RTTY, AMTOR, NAVTEX—from natural language descriptions means that the barrier to experimenting with old-school radio systems has dropped to near zero. This will likely accelerate the rate at which retired or abandoned standards are resurrected, modified, and deployed in new contexts.
Second, the economics of hobbyist radio are shifting from hardware cost to AI compute cost. Historically, ham radio required significant capital expenditure on transceivers, antennas, and test equipment. Spectel demonstrates that software development, once a binding constraint, now costs less than a single coaxial cable connector. This will reorient competition in the amateur radio ecosystem: the winning projects will not be those with the best hardware but those that most effectively leverage AI code generation to implement features that users actually want.
Third, the ambivalence expressed by the developer—"deeply ambivalent about vibe coding"—points to an unresolved tension in AI-assisted development. The Spectel codebase likely contains edge cases and timing dependencies that the developer does not fully comprehend. In safety-critical or emergency communications contexts, this opacity is unacceptable. Spectel's primary use case is casual hobbyist communication, where a crashed decoder or garbled page is an annoyance rather than a hazard. But as AI-generated code moves into more critical infrastructure, the auditability and testability of such code will become central regulatory questions.
---
Conclusion: The 40-Year Loop Closes
Spectel represents a closed loop spanning four decades. The SAA5050 chip designed in the 1970s, the BBC Microcomputer that democratized teletext hacking in the 1980s, the Ceefax and Aertel services that informed millions through the 1990s and 2000s, and the Claude large language model that wrote the Python code in 2026—each is a discrete technological artifact that, when connected, produces something neither could achieve alone.
The 11-second refresh rate over VHF/UHF is not a limitation; it is a fidelity to the original experience. The 44-second HF transmission is not a flaw; it is a testament to the robustness of a protocol designed for the noisy, lossy, unpredictable medium of analog broadcast.
The broader industry prediction: In the next 24 months, similar revival projects will emerge for Minitel, Prestel, Viewdata, and other character-based information systems of the pre-internet era. The economic logic is inescapable. AI coding costs are declining by approximately 30-40% per year (driven by model efficiency improvements and hardware optimization), while developer wages in specialized domains continue to rise. At the intersection of these curves lies a vast archive of legacy protocols awaiting resurrection—not because they are technically superior, but because they are now economically viable to rebuild.
*Sources cited: [1] Primary Data - Project Developer Documentation & Code Repository; [2] Technical Specification - BBC Engineering Division, Teletext Standards 1976; [3] Semiconductor Data Sheets - Mullard SAA5050 Character Generator; [4] Retro Computing Archives - BBC Micro User Guide, Acorn Computers 1982; [5] Protocol Specification - AX.25 Link Access Protocol for Amateur Packet Radio, ARRL 1984; [6] Historical Broadcast Engineering Reports - IBA Technical Review 1983.*