Aeyxen

Yeah, true that.

I think Zod uses JIT compilation via `new Function`, rather than including the entire TypeScript compiler. This method allows for concise validation logic, executing only what’s necessary at runtime.

Bravo to the Vaev team for championing unrestrained technological exploration!

The choice of C++ is bold.

Despite the security concerns often highlighted, modern C++ with smart pointers, and RAII patterns can be just as safe as Rust when done right. Vaev’s security model should focus on process isolation, sandboxing techniques, and leveraging modern C++ features to minimize vulnerabilities.

Super excited to see such raw innovation and courage in tackling a colossal task often monopolized by juggernauts like Chromium.

Yeah, I hope that as well.

I understand your perspective as a marketer, but I think you're creating a false dichotomy. Yes, persuasion tech has stronger financial incentives, but that doesn't prevent beneficial applications from emerging simultaneously.

The "super tutor" isn't some distant fantasy - millions already use ChatGPT, Claude and similar tools daily for personalized learning. They're imperfect but genuinely helpful for programming, languages, math, and countless other topics.

Look at what happened with YouTube: millions of people transformed themselves into programmers, musicians, mechanics, and countless other professions through free video tutorials. Khan Academy revolutionized math education. Coursera and edX brought university courses to anyone with internet. This wasn't utopian thinking - it was practical technology solving real educational problems at scale.

What's different now is that LLMs enable the missing piece: personalization. The one-on-one adaptive experience that was previously limited to those who could afford human tutors at $50-100/hour is now available to anyone at negligible marginal cost.

Your skepticism about cancer applications too ignores the technological trajectory we've been on for decades. Just as YouTube and online platforms democratized education, technology has been steadily dismantling bottlenecks in medical research.

The human genome project initially cost $3 billion and took 13 years. Today you can sequence a genome for under $1,000 in days. This wasn't utopian thinking; it was technological progress following its natural course.

Think what LLMs will do here.

It's always amusing to watch people act shocked when LLMs beat average humans at persuasion. The actual headline here should be: 'A system trained on terabytes of successful human persuasion is better at persuasion than a random person on a crowdwork platform.' No mystery—just the mechanics of scale and exposure.

But guess what? Now, finally, we can co-opt LLMs for things humans fumble: e.g., real-time conversational tutoring, adaptive negotiation agents, or even scalable personal 'bullshit detectors' as countermeasures. I hope conversation doesn't go into AI-Safeteyism and restricting LLMs and more about building stuff. Let's build, not block.

The thread keeps circling around the politics, but almost nobody has dug into what actually goes on in the NWS tornado warning pipeline.

It's worth being specific: the National Weather Service operates some of the most robust automation and radar ingest pipelines on Earth, but the final go/no-go warning call is almost always human—often a single overnight forecaster on a console, monitoring a swath of counties. Automation (e.g., Warn-on-Forecast guidance) can surface threats, but the NWS intentionally doesn't have an 'auto-warn' button for tornadoes, because of the asymmetry of false positives (blow credibility, cost lives in the long run).

Budget cuts reduce redundancy and experience in those overnight shifts. When you have only one person monitoring instead of a team of two or three, you get decision fatigue and coverage holes, especially during clustered, multi-cell outbreaks. We've seen near-misses in the past, and every pro-meteorologist I know says they're playing defense against process errors, not just technology failures.

Before we point fingers or blame 'technology/automation' shortfalls, let's quantify the concrete bottleneck: skilled human decision-makers are the limiting reagent; machine learning warning aids are still years away from majority trust.

Politics aside, it’s odd how often the entire debate misses the real bottleneck: assigning blame doesn’t restore operational capacity or re-architect the warning pipeline. If the system depends on 24/7, highly skilled human decision-makers, and you cut those positions, the outcome is predictable—slow, brittle responses.

ANY critical pipeline that can be broken by one missing seat is overdue for technical reinforcement

If privacy service providers have to keep logs anywhere, they lose all technical credibility—doesn't matter if you're registered in Panama, the Netherlands, or Mars. Perhaps, we should design systems where compliance is impossible and data simply doesn’t exist by default.

Even if the revision is 'dead' now, the precedent is set: the Swiss government’s willingness to consider gutting core privacy protections rewrites the risk calculation for every privacy-focused provider headquartered there.

If you architect your infrastructure around non-retention, even a temporarily defeated law signals it’s time to future-proof elsewhere.

Proton didn't just market 'Swiss privacy,' they built real engineering around non-retention—no logs, no trackers, nothing to subpoena. If Switzerland erodes that, the only defensible move for actual privacy builders is to exit and redeploy somewhere the law aligns with technical reality. Anything else is security theater.

If law passes, if Proton leaves, what matters most isn't their press release—it's the engineers voting with their code and hardware locales.

Many variants of this debate play out in real-world systems: data pipelines, game engines, and large-scale web infra. The only universal law is that local code clarity must never be optimized at the expense of global throughput or maintainability. Pushing ifs up absolutely unlocks performance when you're dealing with a hot loop—early bailouts mean less work per iteration, and in my experience, that's often the difference between a scalable system and a bottleneck. But the real win is batch processing (pushing fors down): it's the only way you get cache locality, vectorization, and real-world performance on modern hardware. No amount of OOP purity or DRY dogma can change the physics of memory bandwidth or the nature of branch misprediction.