Josep Jaume

CTO at Ramensoft, occasional guitar player, relentless builder of stuff.

Latest posts from Josep Jaume's hut

• Forgive me Father, for I have vibed

  Dec 08 ⎯ I won’t try to hide it anymore: I vibe code. I do. A freaking lot. More than I feel comfortable admitting in public. And every time I do, a sort of discomfort creeps in. A sense that I’m doing something morally dubious, like violating an unspoken code of craftsmanship. Because inside me, two forces coexist, and they rarely agree on what “good work” looks like. Like a modern-day Edward Mondrake, I too hear demonic voices whispering in my sleep. The manager The vibes make my manager instincts kick in. This part of me cares about speed. About validating ideas. About time to market. About delegating. The manager in me doesn’t get obsessed about the details. It understands that customers don’t care about code they won’t see. That perfect is the enemy of done. The prompt machine ends. The reasoning storm stops. The manager, arms crossed, looks at the outcome. Good. This is great. I can’t believe I managed to do in 20 minutes what would’ve taken a team of five an entire week. Crazy times. But… Why… Is… that useEffect there??? This is clearly DERIVED STATE!! The artisan The artisan in me cares about quality. About coherence. About beauty, even where no one else will look. About existential reasons first, money later. The crafter looks at the vibed code and chuckles. This is junior code. You’re an old man who has spent two decades in the industry writing code every day. Look at yourself in the mirror. You’re deplorable. Roll up your sleeves and do the fucking work, you lazy bastard. That useEffect gets immediately obliterated. It needed to be derived state, and the variable namings are also weird all over the place. God, this is going to take a while, but it’s worth the effort. Let’s do this. And now that I’m here, maybe I could refactor this other despising block of sheer vibery that crept in somehow. This is good. I feel like myself again. Wheeew. I just spent the entire afternoon in monk mode. And then the manager knocks: Dude, you just invested 4 hours of your time doing NOTHING. Have you seen the roadmap? Step aside, I think I’ll just prompt my way out of this refactor. And so the circle repeats. Endlessly. Precariously. Hilariously. Where this leaves me (us?) I refuse to believe I’m the only poor soul caught in this absurd, self-inflicted torment. Others must be out there, questioning their seniority or skills in the face of technology so powerful it bends reality around it. A technology that’s indistinguishable from magic—but magic from a Terry Pratchett universe (the blue color indicates an intentional em dash). A universe where wizards cast their spells using flimsy wands, mutter half-formed incantations, and lopsided aberrations materialize out of thin air. Here’s your all-powerful dragon! A dragon that flies for five minutes in utter agony before begging you to end its suffering. And the wizard in you just sighs. Because it knows exactly what comes next: fix its malformed wings, reorganize its guts, invent a digestive system it never asked for. Because somehow, unbelievably, that’s what it means to be a programmer in 2025. Forgive me Father, for I vibed. And I will vibe again. Because the magic is flawed, the outcome uncertain. But hey, at least I got myself a fucking dragon.

  

• AI is the opposite of art, so it can't replace it

  Nov 13 ⎯ I was born in the mid-80s and many of my deepest memories come from the early 90s. Back then the future of computers felt clean and certain. We imagined machines would only get better at what they were built for: being tireless, logical, and perfectly factual. The hard part, as most sci-fi movies pictured it, would be making them sound human. And yet here we are, nearly four decades later, talking to machines in ordinary language and getting fluent answers back. It’s as close to a miracle as technology gets, and it naturally stirs something deep and uneasy within us. What nobody foresaw was the flip side of that miracle. We now have systems that can hold a conversation, write poetry, and imitate style— but they err at the very things computers were meant to master! They miscount, hallucinate facts, and are notoriously sycophantic, eager to agree and to please up to the point of being sometimes useless. Hell, we don’t even know if they know what they don’t know. We long assumed the human-like part would be hard and the logical part easy. Instead we got machines that talk like us, complete with our contradictions, but without the rigor we once took for granted! Art is challenging the norm Because of these qualities, when the first generative tools appeared, many feared they would kill art. If anyone can create a painting, a song, or a story in seconds, who will bother struggling with a canvas, chord changes, or empty notebooks? Won’t creativity collapse and artists get depressed? I think that fear misses the point. I won’t try to define what art is — nobody has been able to so far — but I will point out one of its intrinsic properties: artistic expression is risk taking. All we’ve seen in AI so far is rooted in its probabilistic nature, in their ability to autocomplete the next token based on drinking human knowledge from a firehose. By contrast, art is rooted in desire, in the will to make a statement, to want things, to surprise and to impress. This is exactly the opposite of what generative AI is trained to. Where artists are risk taking, AIs are risk averse. They are destined to be mediocre by design! Some people would argue that AIs will just get better and eventually become completely human-like. But here’s the thing - it will take being human to create human art that’s captures the interest of other humans, with all its flaws and imperfections. Take chess, for example. Chess engines have beaten grandmasters since the 1990s, yet human-versus-human chess is thriving. We want to see the drama of risk and failure—the story of a person pushing past their limits. We don’t want to see machines race; we want to watch humans become. Humor shows this vividly. Joking still largely evades language models, because humor relies on cultural context, timing, and a shared understanding of human frailty. My uncle liked to say that joking is the most elevated form of expression, a sure sign of intelligence. We laugh not only at a punch line but at the invisible web of references and double meanings behind it—connections no prediction engine can make. Culture needs intention, not just output So the paradox of our time is that the more machines create, the more we demand to know to which extent humans are involved. What is their story? What do they crave? Is there a world where I could be like them? Far from making us obsolete, we’re increasinly growing more interested in what only people can do. We crave proof of humanity. Music shows it clearly. Streaming has flooded the world with more recorded sound than ever, and yet we’re in a golden age of live music. People don’t gather just to hear songs; they gather for the unrepeatable moment of belonging. Musicians respond by sharing more of their private lives and even their flaws, so we can see the person behind the sound. Economists would say that what we’re experiencing is content inflation. As Jack conte described in his “Death of the follower” talk, the past ten years have been defined by Social Networks shifting from follower first feeds to content-first mindless algorithms. But with myriads of AI-generated content pouring into our doomscrolls, the scarce becomes valuable again. We can’t easily separate the noise from the signal unless we go back to the source: the human behind it. It feels like the pendulum may already be swinging back. Patreon was first to reclaim that space, but we’re seeing similar platforms like Substack repeat the same pattern. This very tiny place in the woods — Fika — is also an attempt to bring back attention to the people in a forest of infinite noise 🌲💚 AI Makes Craft More Valuable You might be thinking: This is very optimistic and all, but what about all the people whose livelihoods are apparently ruined — painters and illustrators, studio musicians, copywriters? Well, history offers perspective. Consider IKEA. A century ago most people bought expensive furniture from local carpenters. Then IKEA arrived with stylish, affordable pieces, and much everyday woodworking disappeared. Yet something unexpected happened: people grew more aware of true craft. Handmade pieces and small wood fairs began to thrive because mass production educated our taste. By making the ordinary cheap and plentiful, IKEA increased the value of the unique. Likewise, AI will likely expand appreciation for genuine artistry, even as it commodifies the mundane. AI is democratizing artistic creation The fact that AI can’t create art by itself doesn’t mean that it can’t enable artistic expression for people. In my own town, a retired school teacher recently staged a full musical without ever learning how to compose. After a lifelong passion for music and theater, he shaped melodies and lyrics using Suno until he had a full 24-piece musical. Isn’t the act of curation also creation? The result is inspiring: kids and adults alike will now participate in the plays and local musicians will get involved in the live act. Heck, even me and my sister are getting involved in recording it. We got a ton of artistic expression worth where none would have happened just of a few years ago! Companies like My Sheet Music transcriptions — good friends of mine — are booming in an era of professional and amateur creators that want to turn their creations into music played at their parties, weddings, or just to be able to silently enjoy by themselves. Heck, doesn’t this send shivers through your spine? AIs as Mass Value extractors With all the optimism of this piece, I’m not unaware of the massive value extraction the likes of OpenAI, Anthropic or Meta are perpetrating by sniffing the whole of internet built by all of us and claiming it as theirs, just to generate billions in revenue and trillions in valuation without giving none of them back to their sources. This is wrong and needs to be fixed. Just recently, Anthropic agreed to pay $1.5bn to book authors, and this is probably just the beginning of more lawsuits to come. Cloudflare is also trying to reshape how robots are able to access content based on consent. We are yet to find a way to find a common understanding that benefits both parties, but we’ll get there. The road ahead This revolution is far from over and will take years, maybe a decade to settle. But AI is here to stay—and I think that is, on balance, a good thing. Every major technological shift, from the printing press to photography to recorded music, has sparked fears of cultural collapse. Yet art always prevailed, because it lives on the frontier of what’s possible, challenging the status quo. Great movements often arise as counter-movements! I for one I’m eager to discover what human ingenuity is capable of in the face of infinite possibility. And here’s the bigger point: the very abundance of machine-made content forces us to become more human, not less. We will gather more often, talk longer, listen harder. We will prize the bold choice, the imperfect note, the surprising sentence. The next artistic revolution will come from people who dare to go where probability cannot follow. Art begins where probability ends. So while AI may be the perfect opposite of art, it might also be the best thing to happen to human creativity. The first time I heard Peter Gabriel’s "Down to Earth" song in the final scene of WALL·E, something clicked: a childlike wonder that humanity—and the will to start again—will prevail.