Hacker News Digest

Friday, September 25, 2026

In This Issue

Hacker News
F-Droid 2.0: A New Chapter for Android Freedom
Show HN: Make cursed fonts like Times New Bastard
Two-Tier Encryption in the UK – Identical Apple Devices, Different Protection
Why is the liver so weirdly regenerative?
Show HN: Whiteboard (YC W26) – An open-source IDE for thoughtful software design
Dutch governments builds alternative for Microsoft based on NixOS
'That's so AI ' What gen Alpha's biggest insult tells us
What Is RLCD? The Secret Behind Jev
Jev Is Not a Language Model, but It Breaks Like One

Zipper Data Brief

September 25, 2026

Your daily digest of the best from Hacker News

1284 points · daveoc64 · comments

# Summary F-Droid 2.0's redesign is appreciated by many users who switched to alternatives like Droid-ify due to poor UI, though some criticize the trendy flat design lacking visual hierarchy. Common concerns include the app's limited catalog compared to Play Store, lack of review systems, and confusion about the rollout timing since many users still see version 1.23.2 as current.

Show HN: Make cursed fonts like Times New Bastard

732 points · MitPitt · comments

# Summary Developers and designers are enthusiastically creating deliberately "cursed" or intentionally broken fonts by mixing typefaces, misaligning baselines, and manipulating glyphs for pranks and artistic effect. The tool has sparked nostalgia for similar font manipulation projects and playful discussions about the chaotic aesthetics these creations produce.

Two-Tier Encryption in the UK – Identical Apple Devices, Different Protection

468 points · ReturnoftheHack · comments

# Summary The discussion centers on Apple's decision to disable Advanced Data Protection (ADP) for UK users due to government pressure, with commenters viewing this as either a capitulation to authoritarian surveillance demands or a reasonable compromise. Critics argue this effectively creates a backdoor for UK law enforcement while others question why Apple doesn't resist more forcefully or whether users can circumvent the restrictions.

Why is the liver so weirdly regenerative?

434 points · jbotz · comments

# Summary The discussion explores why the liver regenerates so well, with commenters noting that evolutionary pressure favored regenerative abilities for organs exposed to toxins and injury rather than perfect whole-body regeneration, while also correcting the article's oversimplifications about neuron regeneration, wound healing, and evolutionary teleology.

Show HN: Whiteboard (YC W26) – An open-source IDE for thoughtful software design

331 points · sidharthkmenon · comments

# Summary Whiteboard generates visual architecture diagrams and semantic diffs from code using AI agents, helping developers understand design decisions—though reviewers question whether it's truly an "IDE" (it can't edit files), worry about diagram accuracy/drift from code, and debate if it's narrow enough to justify adoption over existing tools like Mermaid or custom solutions.

Dutch governments builds alternative for Microsoft based on NixOS

330 points · fjfaase · comments

# Summary European governments (Dutch, French, German) are adopting NixOS-based systems as open-source alternatives to Microsoft, with NixOS praised for its declarative and reproducible approach, though concerns remain about replacing Microsoft Office functionality and managing Linux systems at scale.

AI / Machine Learning

'That's so AI ' What gen Alpha's biggest insult tells us

168 points · YeGoblynQueenne · comments

# Summary Gen Alpha uses "That's so AI" as an insult for low-effort, inauthentic content, reflecting their sophisticated ability to detect slop and AI-generated material—though some commenters debate whether the trend is overstated and whether the phrase conflates all AI use with poor quality.

What Is RLCD? The Secret Behind Jev

66 points · tnspacetime · comments

The discussion clarifies that RLCD (Reinforcement Learning for Calibrated Decisions) addresses the practical problem of LLMs giving inconsistently scaled preference scores by using statistical calibration techniques to normalize results.

Jev Is Not a Language Model, but It Breaks Like One

33 points · patresh · comments

The commenters are skeptical of both the paper's premise and presentation: they question the practical relevance of an attack that requires direct document modification (which would already be game-over for security), and they criticize the authors for using an LLM to write about LLM vulnerabilities rather than explaining their findings directly.

For Computer Use, the harness matters as much as the model

29 points · MiguelG719 · comments

# Summary Browserbase released a new benchmark showing that the choice of harness (framework/wrapper) for browser agents is nearly as important as the model itself, with significant differences in cost and accuracy across nine tested harnesses. The discussion highlights growing interest in specialized harnesses as a competitive advantage and raises questions about tool choices, model selection, and the practical viability of browser agents versus newer computer use models.

AI Safety Is Mostly a Sex Cult in Berkeley

118 points · martythemaniak · comments

# Summary The discussion centers on a now-deleted article alleging the AI safety movement in Berkeley is cult-like and sexually permissive, with commenters split between defending the community's unconventional lifestyle and ideas as harmless innovation, and criticizing it as a distraction from real societal harms of AI automation.

Startups / Business

Show HN: PlaceCall (YC W26) – agentic API to call businesses and get things done

16 points · ymarkov · comments

# Summary Users are asking about market fit in different geographies and whether the API can be used for sales development rep tasks, suggesting practical interest in the tool's business applications.

Anthropic Strikes $12B AI Computing Deal with Akamai

10 points · mfiguiere · comments

# Summary Anthropic signed a $11.6 billion, seven-year computing deal with Akamai for CPU capacity to support AI services, with an additional warrant giving Anthropic equity options in Akamai.

Hacker News Digest

In This Issue

Zipper Data Brief

Top 6 Trending

AI / Machine Learning

Startups / Business

More Stories (37)