Pleopods Weekly #29 — September 25, 2026

This Week on Lobste.rs

Trending topics: security culture vcs law ai

1. What is still exciting in tech? ask culture

submitted by ploum — 153 points (+150 this week) — 204 comments

The thread captures real frustration with tech monetization, but the pushback is worth sitting with—Linux gaming actually works now, niche projects like Marginalia solve problems the big players ignore, and maybe the excitement never left, just got drowned out by doomscrolling.

2. Radicle: Disclosure of Vulnerability in the Network Protocol distributed security vcs

submitted by tuxes — 137 points (+128 this week) — 53 comments

Radicle is ditching its network protocol entirely after discovering it leaks private repo data in plaintext and lets attackers spoof nodes — they're switching to iroh and recommend disabling private repos until the migration lands.

3. What Sun got wrong historical

submitted by j11g — 119 points (+108 this week) — 64 comments

Sun had superior strategy and technology but couldn't execute the basic sales and support mechanics that would have turned those advantages into customers — a cautionary tale about how operational incompetence can kill a company that's winning on paper.

4. You Know GDPR Is Good Based on Who Hates It culture law

submitted by siddhartha_golu — 110 points (+103 this week) — 17 comments

GDPR's unpopularity in US tech circles is itself evidence it works — by making privacy enforceable law instead of a competitive afterthought, it undercuts the surveillance-based business model. The cookie banners we all hate are dark pattern theater, not a GDPR failure, but proof the regulation forced companies to acknowledge what they'd rather hide.

5. Typst makes big strides a11y programming

submitted by isuffix — 117 points (+101 this week) — 20 comments

Variable fonts, MathML export, and bundling for multi-document outputs ship in 0.15 — Typst is moving past LaTeX-replacement rhetoric into actual tooling for scientific work.

6. We Must Create the Shit Machine satire

submitted by FedericoSchonborn — 98 points (+91 this week) — 3 comments

A McSweeney's satire about tech's cycle of inventing problems, selling subscriptions to fix them, then shipping apps that actively make things worse — featuring notably useless smart umbrellas.

7. Everybody's Lost Their Minds culture

submitted by j11g — 193 points (+87 this week) — 63 comments

A working engineer calls out the real cost of the AI hype cycle: companies are pulling senior people off actual security work to chase AI vulnerabilities that don't move the needle, while the industry performs regulatory compliance theater and ignores the environmental damage. The uncomfortable truth is that patching was never the hard part of security—but we've successfully convinced ourselves it is.

8. Flock cameras are riddled with security vulnerabilities and hard-coded credentials reversing security

submitted by robey — 108 points (+86 this week) — 9 comments

A researcher cracked open Flock camera firmware and found it's running on ancient Android with hard-coded provisioning keys and plaintext secrets, raising real questions about how these devices made it into police departments.

9. No Sloptober ai

submitted by jeremiahlee — 89 points (+85 this week) — 25 comments

A month without LLMs to see what skills have atrophied and whether the productivity gain was worth the cognitive cost—the author suspects you'll be surprised at what you've forgotten how to do solo.

10. AI Is an Elite Crime Spree law

submitted by jcd — 93 points (+83 this week) — 33 comments

The author argues that AI companies are committing mass copyright infringement and treating legal liability as a business expense to be absorbed if caught — and that the current enforcement approach amounts to letting them off the hook.

11. Looking forward to Git 2.56 - and 3.0 vcs

submitted by dimonomid — 85 points (+75 this week) — 17 comments

Git 2.56 ships 700+ commits of polish, but the real intrigue is whether it's the last of the 2.x line—a GitHub engineer just signaled SHA-256 support is close enough that jumping to 3.0 might actually happen.

12. Beyond jj: config & tools ecosystem vcs

submitted by abhin4v — 85 points (+74 this week) — 3 comments

Jj's capabilities have grown past the point of needing external scripts for common tasks — jj converge handles automatic reconciliation, jj bookmark advance eliminates the old workaround pattern, and there's now a whole catalog of native commands and config tricks people have built up. The ecosystem piece maps what actually works in practice versus what still requires wrapper scripts.

13. The scourge of x86 emulation retrocomputing

submitted by videah — 74 points (+66 this week) — 0 comments

The fundamental mismatch between x86's strict memory ordering and ARM's weak model forces emulators to turn nearly every load and store into expensive acquire/release instructions — a performance hit that ranges from barely noticeable on modern cores to severe on others like AmpereOne.

14. Named and Optional Arguments are Awesome plt rust

submitted by jado — 66 points (+63 this week) — 39 comments

Rust's lack of named and optional arguments forces the standard library to expose eight different HashMap constructors where one would do, and makes it easy to accidentally swap boolean parameters—the author walks through how other languages solved this and what Rust could steal from them.

15. Be alert: targeted attacks on prominent Rustaceans rust security

submitted by itamarst — 97 points (+62 this week) — 32 comments

Attackers are using fake video calls to target Rust maintainers with malware and compromised commands, reusing tactics from the arrayref supply-chain attack last month.

16. I don't like passkeys security

submitted by eduard — 70 points (+58 this week) — 46 comments

Passkeys eliminate phishing but swap one problem for others: permanent lockouts if you lose your device, lock-in to Apple or Google, and a fragmented ecosystem that can't yet match what password managers already do—making them enterprise security theater for most people today.

17. I Built Non-Autoregressive Decision Models a Year Ago. Then a Frontier Lab Called It a "Breakthrough" ai

submitted by Yogthos — 59 points (+58 this week) — 6 comments

A year ago this engineer published non-autoregressive decision models; when a well-funded lab recently claimed the same approach as a breakthrough, he rebuilt it open-source and cut latency from 150ms down to 33–38ms.

18. Thoughts on the Future of Web Browsers browsers

submitted by orib — 66 points (+57 this week) — 29 comments

The author removed over 850K lines of AI code from Firefox and argues that forks need to delete these features outright rather than just disable them—the integration is getting too deep to patch out later, and coordinated maintenance across forks probably isn't sustainable anyway.

19. Breaking Up with Google Play: Why Conversations Is Now Free android

submitted by tusharhero — 61 points (+57 this week) — 4 comments

After 12 years of treating the Play Store as his rent money despite chronic friction with Google's review process and unresponsive support, Daniel Gultsch finally has enough grant funding to leave — and he's making Conversations free and directing users to F-Droid instead.

20. ChatGPT now knows what you do on other websites via ad collector privacy

submitted by jrgtt — 60 points (+56 this week) — 7 comments

OpenAI's ad pixel tracks your browsing across 936+ advertiser sites and links it to your ChatGPT account even when you've refused marketing consent—the researcher reverse-engineered the whole flow and found it also exfiltrates unencrypted postal codes and page paths from medical and other sensitive sites.

Pleopods is a weekly digest of the top links from Lobste.rs. Unsubscribe | Archive