--- title: "The Reckoning" url: https://www.localghost.ai/hard-truths/reckoning description: "Is the rot inevitable? How do ethical builders survive? What happens when a generation never knew privacy? Hard questions, honest answers." updated: 2026-10-24 site: LocalGhost.ai index: https://www.localghost.ai/llms.txt --- > HARD_TRUTHS // POST_02 // JAN 2026 # The Reckoning [Vlad Cealicu](https://www.localghost.ai/about#vlad-cealicu) // 02 January 2026 Writing [manifestos](https://www.localghost.ai/manifesto) is easy. Drawing [timelines](https://www.localghost.ai/hard-truths/inflection) is easier. But manifestos don't answer the hard questions. This page tries to. A good friend read the manifesto and we ended up talking for hours. She had to google some words, she pushed back on the parts that sounded too easy, and she asked the questions I'd been avoiding. *Is the rot inevitable? How do builders who don't sell out survive? What happens when a whole generation never experienced privacy in the first place? And if AI is a yes-man trained on human sycophancy, isn't that its own quiet catastrophe?* The conversation was too useful to keep private, so here it is, her questions, my best attempts at answers, the uncomfortable parts included. Some of what follows is signal and some is alarm, and the labels are honest about which is which. ## 1. IS THE ROT INEVITABLE? Every platform follows the same arc. First they delight you to get you hooked, then they squeeze you to pay back investors, then they strip the copper from the walls to hit quarterly numbers. Cory Doctorow named it enshittification, and the business model is working as designed. 01 YEAR 1-3 Delight Users 02 YEAR 4-7 Extract Value 03 YEAR 8+ Strip the Copper > [ALARM] Every company answering to shareholders eventually answers only to shareholders, and the product becomes a mechanism for that rather than a reason for existing. But that's fatalism about companies, and code doesn't follow the same arc. A public repository can't be ruined in a board meeting, it can die from neglect but it can't be deliberately made worse, and anyone can fork it and keep going when the original maintainer stops caring or gets acquired. > [SIGNAL] The rot is inevitable for the company wrapped around the technology, and the defence against it is making sure the technology can survive without the company, which is what open source exists to do. ## 2. HOW DO ETHICAL BUILDERS SURVIVE? This is the question that quietly kills most privacy-respecting projects, and we don't have great solutions for it yet. You're competing against companies that give their product away for free because you are what they're selling, and when the competitor's product costs users nothing except their privacy it's harder to charge money for yours. WHAT DOESN'T WORK Advertising (the whole point is to avoid tracking people). Subscriptions without lock-in (too easy to cancel). "Free tier with paid upgrades" (race to the bottom). Donations alone (works for a few, not most). WHAT SOMETIMES WORKS Grants from foundations (but they run out). Consulting services around the software (doesn't scale). Hardware sales with a margin (one-time money). Paid support for businesses (small market). WHAT MIGHT WORK New funding models like quadratic funding (still experimental). AI dramatically cutting the cost of building things (happening now). Bounty systems where people pay for features they want (fragmented but promising). > [ALARM] The economics are structurally hostile, and nobody has a clean answer for how to fund software that refuses to monetise its users. But AI is dramatically cutting the cost of building, and one person with good judgment and the right tools can now build what used to require a team. Smaller teams need less capital, and less capital means no investors, so there's no exit pressure and no reason to eventually turn on your users to pay anyone back. [POST_01](https://www.localghost.ai/hard-truths/inflection) mapped the window. The same hardware shift that threatens privacy also makes ethical building viable at small scale for the first time. ## 3. WHAT IF THEY NEVER KNEW? My friend pointed out something that stuck with me, that teenagers today might not understand what privacy feels like in a bodily sense because they've never been unreachable, never had hours pass without anyone knowing where they were or what they were looking at. 1990 BORN BEFORE THE FEED Remembers the internet before Facebook, had a childhood without smartphones, and privacy is a memory of how things used to feel rather than an abstract idea. 2010 BORN INTO THE STREAM "Privacy settings" is just a menu in an app rather than a state of being, and being truly unreachable or untracked is something that has never happened to them. > [SIGNAL] People don't need to understand privacy intellectually to feel when something is wrong, and the discomfort is already widespread even among people who've never known anything different. They feel it when the same ad follows them across every website, when their phone suggests something they only mentioned out loud, when social media keeps showing them content that makes them feel worse about themselves and they can't stop scrolling anyway. The job is to build the exit before they go looking for it and find nothing. ## 4. WHY YOUR AI AGREES WITH YOU Mainstream AI assistants agree with you, validate your reasoning, praise your ideas, and almost never tell you that your plan has obvious holes or that you're lying to yourself, because honesty that makes you uncomfortable is indistinguishable from a bad experience in the metrics they're optimised for, and [the incentive structure that produces this](https://www.localghost.ai/hard-truths/dictator-brain) is one of the harder problems we're trying to solve with LocalGhost. When you ask a friend for advice, they know things you didn't tell them, they remember the last time you said the same thing and didn't follow through, they notice the parts of the story you keep leaving out, and they care about your actual outcome rather than whether you end the conversation satisfied. An AI only gets what you type, filtered through however you want to see yourself. The architecture that keeps your data private, local storage, local inference, no cloud sync, is the same architecture that lets an AI build enough context about you over time to be honest with you, because it knows you and it has no business reason to flatter you. The [manifesto](https://www.localghost.ai/manifesto) covers why privacy and honest AI turn out to be the same engineering problem. ## 5. WHAT HAPPENS WHEN IT FINDS YOUR GRIEF My sister-in-law went through a bad breakup. She talked to ChatGPT about it, the free version, and within days her Instagram had filled up with content about how men are trash, how to move on fast, how relationships are doomed. The algorithm doesn't know the difference between helping someone heal and helping them spiral, it just knows what keeps them scrolling, and grief keeps people scrolling for a long time. > [ALARM] Targeted content is the deeper manipulation because it shapes what you think is normal, what you think is possible, and what you think you deserve, and it's going to get worse as AI-generated influencer personalities optimised purely for engagement replace real people in the feed over the next few years. ## 6. THIS MIGHT NOT WORK Manifestos are optimistic by nature, but privacy-focused alternatives have lost every major battle so far (email, social networks, messaging, cloud storage, phone operating systems). Every time, the convenient default won while the principled alternative stayed niche and got called idealistic. > [ALARM] We might be too late, and the defaults might already be set in ways that matter, and I'm building anyway because the alternative is to watch it happen and do nothing, but I want to be clear that I'm not certain this works. The Cypherpunks didn't win either, not completely, but Bitcoin, Signal and the encryption protecting this page all exist because of them. They built tools that outlasted their movement and gave everyone who came after something real to build on. > [SIGNAL] The goal is to make the exit real enough that it changes how power behaves, which is a lower bar than winning and a higher bar than just making noise. **Strip away the manifesto language and this is what's left.** Building alternatives is worth doing even if they don't win. Because some people will use them, and they matter. Because building in the open creates knowledge that outlasts any single project, and because the questions are hard and anyone telling you they have clean answers is selling something. --- My friend asked whether any of this would work, and I told her I didn't know, and she said that was the first honest thing I'd said all evening. [[ localghost.ai // hard-truths ]] --- Cite this page as https://www.localghost.ai/hard-truths/reckoning. Facts about LocalGhost are kept current at https://www.localghost.ai/about, and the index of the whole site for language models is https://www.localghost.ai/llms.txt.