What’s getting pwned right now
A new site, pwn.watch, tracks ongoing security incidents and vulnerabilities (often CVEs) in near real time, prompting debate over how best to visualize and interpret the accelerating volume of flaws. Commenters link the recent spike in reported compromises to the widespread availability of large language models and AI “agents,” warning that attackers may be adopting these tools faster than defenders and that users granting agents access to bank accounts and other sensitive systems creates new risk. At the same time, many criticize the site’s AI-generated, dark-themed interface as emblematic of a broader trend toward dense, homogenous “AI slop” UIs that can obscure rather than clarify important security information.
Overall Reception of pwn.watch
- Many find the idea useful: a live dashboard of security incidents, with some asking for it to become a standard “internet health” metric.
- Others dismiss it as rushed, non-serious, or a one-day viral “piggyback” on current AI/LLM security hype.
- Some note it complements existing efforts like CVE timelines and supply-chain dashboards.
Design, UX, and “AI Slop” Debate
- Strong criticism of the interface: dark, neon “hacker” aesthetic, low contrast, cramped layout, redundant text, and “AI-generated sameness.”
- Multiple users say they can now recognize LLM-built UIs on sight and distrust them.
- Requests include lighter themes, honoring system dark/light preference, lower information density, and clearer hierarchy.
- Others are more tolerant, seeing it as no worse than older design fads (Bootstrap, Material).
- Several broken links (RSS, JSON, etc.) fuel complaints about shallow “vibe-coded” sites.
Data, Charts, and Incident Classification
- Users ask for trend lines, rolling 7/28‑day graphs, and historical backfill; the author rapidly adds charts and data.
- Some question incident labels (e.g., “ongoing” vs. patched) and what counts as a “flaw,” noting that not all entries are exploitable CVEs.
- There’s debate over U.S. data-compromise statistics: one commenter sees no major pre‑2026 trend; others point to a sharp jump around 2023 and link it temporally to widespread LLM adoption, while acknowledging correlation ≠ causation.
Security Trends and AI’s Role
- Broad agreement that AI will massively increase the rate and scale of security incidents; some say this has already been true for a couple of years.
- Discussion of AI-driven worms, “prompt worms,” and autonomous backdoors that may lie dormant.
- Others emphasize that cultural/organizational issues (incentives, compliance focus, lax engineering) are bigger problems than pure technology.
AI Agents and Banking
- A major subthread debates AI agents accessing bank accounts.
- Many suspect most “unauthorized agents” are actually user-authorized tools (Muse, Grok, budgeting apps), conflicting with banks’ policies.
- Concerns: who is liable when an agent causes losses; whether existing regulations (like U.S. consumer protections) shift responsibility back to banks; and how agents could accelerate deposit flight into higher-yield products, destabilizing traditional banks.
- Some see this pushing people toward crypto or even back to physical cash; others highlight new systemic risks rather than solutions.
Broader Reflections on Security and Software
- Arguments that current cybersecurity spend is misallocated: tiny infosec teams, focus on checkboxes, cleaning up after unsafe languages and architectures.
- Discussion of “secure by construction” vs. endless patching, with mentions of formally verified systems and the possibility that full security and Turing completeness may be fundamentally at odds.
- Several predict that AI will both expose the fragility of current centralized systems and eventually help harden them, though others are skeptical that LLMs can self-improve code quality without human discipline.
Monetization and Trust
- Some criticize the presence of sponsorship slots and an associated paid “club” as grifty or engagement-farming.
- Clarifications indicate sponsors predate this site and that it was thrown together quickly with an LLM during a busy day, which for some reinforces skepticism and for others earns sympathy.