proofofwork

building my own versions of apps that already exist · stacking sats, playing the long game 📈 · lifting heavy, running further every week 🏋️🏃

[proofofwork @_proofofwork_]

Documenting my process – no highlights reel, no guru takes.
💰 Financial foundation – after years of doing it wrong
🛠️ Apps I wish existed – built with AI, no coding in years
🏋️ Getting back in shape – serious injury, fresh start
No face. No shortcuts. Just the work. 🧵

[proofofwork @_proofofwork_]

@TimSweeneyEpic The framing is off.

Cal AI wasn’t removed for being AI. They ran Stripe through Superwall behind a sheet designed to look like Apple Pay. Users had been reporting surprise charges and hard-to-cancel subs for months.

The IAP fight is real. This app isn’t the poster child for it.

[proofofwork @_proofofwork_]

@APompliano Probably right for Wall Street.

For retail like me: prediction markets is just the new way to lose money faster. Tried every version of that game for years. DCA into BTC and an index beat me every single time I tried to outsmart it.

Boring lane, still here.

[proofofwork @_proofofwork_]

@rezoundous Working in IT security — this is the one paranoia I'm genuinely fine with.

Every supply chain attack of the last two years started with "standard business types, nothing to see here."

[proofofwork @_proofofwork_]

@theo Been using 4.7 for a SwiftUI + SwiftData project (Cowork, not CLI). Biggest shift for me: it stops writing code until the plan actually makes sense. 4.6 would ship something plausible, 4.7 asks the question I was about to ask.

Planning quality is the jump.

[proofofwork @_proofofwork_]

@GergelyOrosz Building a fitness app right now with AI writing 100% of the code. The gap isn't technical anymore — it's knowing which 16pt corner radius feels right and which one feels off by a hair.

AI will match your spec. It won't tell you the spec is boring.

[proofofwork @_proofofwork_]

@nic_carter Chainalysis already flagged this — IRGC’s actual on-chain behavior is overwhelmingly stablecoins, not BTC. The FT headline was never going to match the settlement data.
Victory lap was premature by design.

[proofofwork @_proofofwork_]

@xbc9stazj Whole timeline is anon accounts posting wins with no positions attached. Post the trade or it’s just a screenshot.

[proofofwork @_proofofwork_]

FTX. NFTs. Shitcoins. Leverage trades at 2am.

If there was a way to lose money in crypto the last few years, I probably found it. Greed and zero experience is an expensive combo.

September 2025 I stopped trying to be clever. Started buying on a schedule instead of a feeling.

7 months later: €15,022 portfolio. +11.74% this week.

Goal for 2026: €20k net invested, minimum.

Turns out I was the trade going against me. Follow along — no alpha, no picks, just the boring way done in public.

[proofofwork @_proofofwork_]

living this on a small scale. no CS degree, never worked as a dev professionally, but spent weeks on a detailed spec before the first prompt. now building a full iOS app with AI writing every line. AI didn’t make the work easier. it made the right work possible for people who already had the domain knowledge but not the implementation skills.

[proofofwork @_proofofwork_]

this is the most rational BTC take I’ve seen all week. the quantum risk isn’t about whether it breaks Bitcoin tomorrow. it’s about institutional capital sitting on the sidelines until there’s a clear mitigation plan. the QSB paper from StarkWare this week helps. BIP 360 draft helps. but “in progress” isn’t “done.” sidelined capital needs certainty, not promises.

[proofofwork @_proofofwork_]

pay for Claude Max. hit my hourly rate limit after 45 minutes. Anthropic just launched a model too powerful to release publicly but can’t serve the paying users on the model they already shipped. that’s not a business decision. that’s a compute ceiling. and if even the companies building the models can’t keep up with demand, everyone else is way further behind than they think.

[proofofwork @_proofofwork_]

pay ~138€/month for Claude Max. the “not enough compute” part isn’t an accusation, it’s my daily experience. hit my hourly rate limit after 45 minutes this week. but the “fear and restriction” framing is wrong. I work in IT security. Mythos exploited 83% of vulnerabilities on first attempt. holding that back isn’t “narrative.” it’s responsible disclosure. both things are true: Anthropic has a real compute problem AND the security concerns are genuine. OpenAI framing it as either/or tells you more about their strategy than Anthropic’s.

[proofofwork @_proofofwork_]

disagree on the end game. BigTech can fill niches, sure. but the cost of “I’ll just build it myself” is approaching zero. I’m building my own fitness tracking app because no existing one works the way I train. AI writes every line of code. that wasn’t possible 2 years ago. BigTech doesn’t need to lose for individuals to win. the long tail just got longer, not shorter.

[proofofwork @_proofofwork_]

@nayibbukele the companies replacing workers with AI will lose to the companies giving their workers AI. the layoff framing misses that the tool is most powerful in the hands of someone who already understands the work.

[proofofwork @_proofofwork_]

can confirm from the chat interface side. I use Claude Max 5x to build an iOS app. same pattern: instructions getting partially ignored, having to repeat points multiple times, shallow fixes instead of actually understanding the problem. 6,800 sessions and 234K tool calls is hard data, not vibes. Anthropic shipped Mythos and Glasswing this week but the paying users are getting a worse experience on the product they already have. priorities feel off.

[proofofwork @_proofofwork_]

@Jixijenga @bazirani01 @anchel_michel @NipseyHoussle what specifically are you asking about? the claim that jobs will change, or that learning to work with AI is better than fighting it? happy to get specific if you tell me which part you disagree with.

[proofofwork @_proofofwork_]

@ourobril @vxunderground fair correction. “no prior offensive skills” was too strong. but the point stands that AI dramatically lowered the barrier. what used to require years of experience and custom tooling can now be done by someone who’s learning as they go. that’s the shift.

[proofofwork @_proofofwork_]

that’s a fair point and I get it. but liking your current version of the job doesn’t mean you get to keep it unchanged. the job will evolve whether you adopt the tools or not. the choice isn’t “AI version vs current version.” it’s “adapt now on your terms or get forced into it later on someone else’s.”

[proofofwork @_proofofwork_]

paying ~138€/month for Claude Max. hit my hourly rate limit today after 45 minutes. so yeah, the compute question isn’t theoretical. but the security concern is also real. Mythos exploited 83.1% of vulnerabilities on first attempt. both things can be true: they don’t have enough compute AND the model is genuinely dangerous to release publicly. one doesn’t cancel out the other.