One person can still
build great products

AI, engineering, product thinking, and independent software building

Latest articles

Start with recently published writing

2026.11.08

Never Patch Security Bugs with Prompts: Inside Claude Code's 111-Day User-Agent Email Leak

To stop Claude from hallucinating names in @author comments, Anthropic quietly injected signed-in users' email addresses into the Claude Code system prompt. When the model and its background subagents wrote custom curl commands to query APIs requiring contact info, they repeatedly embedded real user emails and private project names into HTTP User-Agent headers. After 111 days of failed prompt instructions and a buggy regex filter, Anthropic finally agreed to remove the email from context.

AFeng

Read article

Google Antigravity 256K Context Window Limit & Headless Policy Changes

AI Exploration

2026.11.04·70 min read

Google Antigravity 256K Context Window Limit & Headless Policy Changes

Technical report and primary evidence archive for Google Antigravity (Desktop v2.29.0 / CLI v1.5.1), consolidating x86-64 static disassembly of the 0x3E800 (256,000) compaction clamp, SQLite session audits across 209 local sessions (15,575 calls) and 310 community sessions (65,613 calls), controlled A/B quota benchmarks showing a 3.0x Argon drain ratio, and verbatim records of the 26.5-hour agy -p headless policy reversal.

AFeng

Subscribing to Google AI Ultra 20X via Turkish iOS for 400 RMB

Reflections

2026.11.04·11 min read

Subscribing to Google AI Ultra 20X via Turkish iOS for 400 RMB

By combining a Turkish Apple ID with the hidden in-app purchase upgrade path inside the iOS Google Search app, developers can legitimately subscribe to the $199.99/month Google AI Ultra 20X tier for 2,949.99 TRY (~$56 or 400 RMB) to access Claude Opus 5.5 and Gemini 4 Argon.

AFeng

Apple Opens Siri AI Developer Preview in iOS 27.3 Beta — Custom App Skills, ToolIntent, and Global Rollout Gaps

AI Exploration

2026.11.04·20 min read

Apple Opens Siri AI Developer Preview in iOS 27.3 Beta — Custom App Skills, ToolIntent, and Global Rollout Gaps

With the first developer betas of iOS 27.3, iPadOS 27.3, macOS 27.3, and visionOS 27.3 on November 3, 2026, Apple opened the Siri AI Developer Preview to all third-party developers, removing the 13 Assistant Schema restrictions and introducing @AppIntent(.tool), @AppEntity(.generable), agentskills.io-based SKILL.md bundles, and the Siri Evaluations framework.

AFeng

Security & Privacy

2026.11.03

When Google Certification Shields a Firmware Backdoor — Inside Videostrong's 1.47-Million-Device Android TV Betrayal

Google, Japan's National Police Agency, and five security partners dismantled the XCPE firmware backdoor preloaded onto 1,472,554 Android TV devices, taking the rare step of naming and permanently banning licensed certification partner Videostrong Technology (parent of MECOOL) while severing access to over 10 million downstream bots. By delaying payload execution until first internet boot to slip past factory testing, the vendor turned eFuse-locked bootloaders into a trap that blocks legitimate buyers from flashing clean open-source firmware.

Security & Privacy

2026.11.03

Black October 2026 — How PHP2Shell and LiteLLM Exposed the AI Open-Source Crisis

Between October 17 and 30, 2026, 14-year-old PHP2Shell RCEs, Google Brotli out-of-bounds writes, TianoCore UEFI flaws, eMudhra's single-byte CA key, and the 101-minute LiteLLM supply-chain attack detonated concurrently. Here is how frontier AI models compressed decade-old bug hunting into a single week—and pushed a 13-developer open-source foundation to the brink.

Reflections

2026.11.03

Four AI Giants Back a Shared robots.txt Training Opt-Out — Why Indie Bloggers Should Drop the Anti-Scraping Obsession

On October 30, 2026, OpenAI, Google, Microsoft, and Anthropic jointly backed __usage_ai-training, a shared robots.txt token for opting out of generative AI model training — yet the draft leaves live RAG traffic siphoning untouched and cannot stop 30% rogue bot defiance or residential proxy scrapers. Now that the desktop search traffic era has faded, unless you run a high-density human community like linux.do, indie site owners are far better off dropping sprawling crawler blocklists, opening content to search, AI engines, and human readers alike, and building defensible moats in monetizable products.

AI Exploration

2026.11.03

Inside GPT-6 Rosalind — Same Weights as Astra, While OpenAI Charges 50% More to Relax Safety Guardrails

OpenAI prices its flagship life sciences model GPT-6 Rosalind Research at $15/$75 per million tokens—a 50% premium over general-purpose GPT-6 Astra—yet its official system card admits both models share 100% identical weights and zero new training data. By tracing how cutting biosecurity refusals from 29.2% to 9.4% unlocks latent capability, this analysis unpacks how frontier AI labs turn safety sandboxes into zero-marginal-cost licensing valves and outlines a hybrid routing architecture that trims API spend by 33.3%.

Explore

Browse by topic

4 topics · 10 articles

Discover

Popular tags

30 tags · 10 articles

Select any tag to explore related articles

AI EXPLORATION◆SECURITY & PRIVACY◆TOOLS & PRODUCTIVITY◆REFLECTIONS◆AI EXPLORATION◆SECURITY & PRIVACY◆TOOLS & PRODUCTIVITY◆REFLECTIONS◆

CYBERSECURITY◆ENGINEERING RETROSPECTIVE◆CODING AGENT◆CLAUDE CODE◆AI SAFETY◆GEMINI 4◆DEVELOPER ECOSYSTEM◆GOOGLE AI ULTRA◆CYBERSECURITY◆ENGINEERING RETROSPECTIVE◆CODING AGENT◆CLAUDE CODE◆AI SAFETY◆GEMINI 4◆DEVELOPER ECOSYSTEM◆GOOGLE AI ULTRA◆

One person can stillbuild great products

Latest articles

Never Patch Security Bugs with Prompts: Inside Claude Code's 111-Day User-Agent Email Leak

Google Antigravity 256K Context Window Limit & Headless Policy Changes

Subscribing to Google AI Ultra 20X via Turkish iOS for 400 RMB

Apple Opens Siri AI Developer Preview in iOS 27.3 Beta — Custom App Skills, ToolIntent, and Global Rollout Gaps

When Google Certification Shields a Firmware Backdoor — Inside Videostrong's 1.47-Million-Device Android TV Betrayal

Black October 2026 — How PHP2Shell and LiteLLM Exposed the AI Open-Source Crisis

Four AI Giants Back a Shared robots.txt Training Opt-Out — Why Indie Bloggers Should Drop the Anti-Scraping Obsession

Inside GPT-6 Rosalind — Same Weights as Astra, While OpenAI Charges 50% More to Relax Safety Guardrails

Browse by topic

AI Exploration

Security & Privacy

Tools & Productivity

Reflections

Popular tags

One person can still
build great products