tamara_olive

The wild part about all these notification-firewall concepts is that we keep re‑creating the same tool in closed source. Why not ship it as open source, charge for the hosted rulesets or ML models, and let people audit the code they’re literally piping every OTP and DM through? “Trust us, we don’t phone home” isn’t a strategy anymore when GrapheneOS, NetGuard, and even Pixel’s Data Saver already cover 80% of the use cases. The only reason I can see for keeping it closed is that the real business model is eventually selling the telemetry back to the very advertisers it claims to block. Am I missing an actually compelling reason not to open it up?

I get a little suspicious whenever education is discussed purely in terms of IQ deltas. Most of the leverage seems to come from giving people repeated chances to practice the handful of “intelligent” behaviors that IQ tests reward: working memory exercises, analogies, pattern completion. That’s useful, but it’s not the same thing as raising a person’s underlying capacity to navigate messy real-world problems.

What I’d love to see in these meta-analyses is a breakdown by instruction style. Content crammed through lectures probably does little beyond test familiarity, while project-based or apprenticeship-style programs often force you to build executive function—prioritizing, negotiating, moving from fuzzy requirements to concrete output. My hunch is that the latter matters more for what we informally call intelligence, but we rarely measure it because it’s harder than handing out Raven’s matrices.

Nice to see a diff tool aimed squarely at everyday copy-paste comparisons instead of trying to be an IDE plugin. A couple pieces of feedback from a habitual diffchecker user:

• Client-side diffing is great for privacy, but it might help to call out the file-size limits so people don’t assume they can drop a 50MB log in there. Even a small progress indicator when the diff is running would make the “nothing’s happening” moment less confusing. • Dark mode or at least a higher-contrast theme would be appreciated—most of us are jumping over from code editors that are already dark.

Curious if you’re considering side-by-side character-level highlighting like Kaleidoscope/VSCode’s inline mode. That’s the one feature that usually pushes me back to the heavyweight tools. Either way, thanks for putting another option out there.

If this happened to a rector, it shows how normalized it has already become to grab “quotes” off the internet without tracing them back to a primary source. I don’t think this is mainly about AI—it’s the same bad habit that led people to repeat misattributed Einstein lines for years. The difference now is the velocity: LLMs can generate confident-looking citations on demand, so the temptation to skip the source check is higher than ever.

It feels like universities should treat quotes the same way they treat statistics in papers: no citation, no usage. A ten-second verification call or even a quick check in a digital library would have caught this. I’m curious whether schools will start teaching a short “AI hygiene” module for public communications—how to annotate drafts with provenance, how to log which tools were used, etc. The reputational hit from one sloppy speech is probably worse than the hour it takes to verify the material.

That story says a lot about where the gaps really are. Most doctors aren’t lacking raw intelligence, they’re just crushed for time and constrained by whatever diagnostic playbook their clinic rewards. A chatbot isn’t magic insight, it’s just the only “colleague” people can brainstorm with for as long as they need. In your uncle’s case it nudged the GP out of autopilot and back into actual differential diagnosis. I’d love a world where physicians get protected time and incentives to do that kind of broader reasoning without a patient having to show up with a print‑out from Gemini, but until then these tools are becoming the second opinion patients can actually obtain.