A ten-year-old XKCD comic about mundane-looking tasks has resurfaced because the software world around it has changed so dramatically. What once looked like a tongue-in-cheek jab at human labour now reads as a marker of how quickly machine vision has matured.
The source material in the packet is a short post by Simon Willison marking the anniversary of XKCD 1425, “Tasks.” Its point is not elaborate: one of the comic’s classic prompts, “check whether the photo is of a bird,” has gone from something that once felt highly specialised to something that can now be handled trivially by a vision large language model. That is a small sentence with a large implication. Tasks that once served as examples of difficult perception problems are now routine demonstrations of model capability.
The significance is less about the comic itself than about the speed of the shift. XKCD has long been a reference point for technical readers because it distills computer-science ideas into memorable scenarios. In this case, the comic’s joke survives, but the technology context surrounding it has changed enough that the joke lands differently. What used to stand in for a near-impossible task can now be automated with consumer-facing tools.
That matters for how engineers and product teams think about benchmarks. If a problem is solved by a model in seconds, it can no longer be used as a useful marker of frontier capability in the same way. The broader lesson is that capability thresholds do not stand still. A good benchmark can become stale very quickly, especially in areas such as vision, language and multimodal reasoning where model performance improves in visible jumps.
The anniversary also serves as a reminder that technical culture often tracks progress through memes, comics and side observations rather than through formal reports alone. A comic about tasks becomes a time capsule when the thing it joked about has been turned into a commodity feature. That is why the post resonates: it marks the point at which a once-difficult classification problem can be demonstrated almost casually.
The source does not provide broader research claims or hard benchmark figures, so those should not be invented here. It simply uses the comic’s anniversary to point to a real shift in what modern vision systems can do. That makes the post valuable as commentary on the pace of change in AI, not as proof of any single model or product milestone.
For readers outside the machine-learning bubble, the takeaway is easy to grasp. Ten years is a long time in software. In this case, it is long enough for a joke about a hard perception task to become a joke about a solved one.
