Threat intelligence editor

Alex Kim

Articles: 235
Desks: 7

Focuses on cloud identity, incident response, and exploitation trends.

CloudIncident responseIdentity

Coverage by Alex Kim

Data graphic: Claude Fable 5 access by plan from July 20, 2026. Max subscribers can spend up to 50% of their weekly limits on Fable 5 at no extra cost, from the same shared pool. Monthly prices drawn to scale are Free $0 no Fable 5 , Pro $20 credits only , Max 5x $100 and Max 20x $200 both included, about 5x and 20x Pro per-session usage . Fable 5 API rates are $10 input and $50 output per million tokens, so Pro's one-time $100 credit buys about 2 million output tokens.

Claude Max Subscription: Is It Worth It Now That Fable 5 Is a Permanent Fixture?

Anthropic made Claude Fable 5 a permanent part of its Max plan on July 20, 2026. See who benefits, who loses access, and whether Max is worth the price now.

Jul 21 · 6 min

Data graphic: Anthropic's vendor-supplied Fable 5 launch table on SWE-Bench Pro, where Fable 5 scores 80.3% against Opus 4.8 at 69.2%, GPT-5.5 at 58.6% and Gemini 3.1 Pro at 54.2%, while the rumored Opus 5 has no published score and is only said to match Fable 5, not beat it. Side panels show Fable 5's 11.1-point lead over Opus 4.8, Humanity's Last Exam 59% vs 41.4% and 44.4% and FrontierCode 29.3% vs 5.7% results, an AA Intelligence Index of 60 for Fable 5 vs 59 for GPT-5.6 Sol, and the leak's unverified half-price claim.

stableResearch Anthropic Opus 5

Anthropic's Own Benchmark: Why the Rumored "Opus 5" Is Being Aimed at Matching — Not Beating — Its Flagship Fable 5

Leaks claim Anthropic's rumored "Opus 5" will match but not beat its flagship Fable 5. We scrutinize the unconfirmed benchmarks and what it means for buyers.

Jul 21 · 6 min

Data graphic: Kimi K3 vs Claude and Codex. Moonshot AI's Kimi K3 has about 2.8 trillion parameters, making it the largest open-weight model, with 16 of its 896 experts active per token. Independent evaluators rank it first on Arena for building web interfaces, second overall on Vals AI behind Fable 5, comparable to GPT-5.5 and Opus 4.8 on Artificial Analysis, and only ninth on Arena for broad text queries. It has a 1 million-token context window, costs about 3 to 4 times more per token than the older Kimi K2 series, and was released July 17, 2026. The verdict: test K3 for frontend work, but keep Claude or Codex as primary for hard, multi-file engineering.

stableResearch Kimi K3

Kimi K3 vs. Claude and Codex: Can China's Biggest Open Model Really Replace Your Coding Assistant?

Moonshot AI's Kimi K3, the world's largest open-weight model at 2.8T parameters, tops WebDev benchmarks but can it really replace Claude and Codex? A team-by-team verdict.

Jul 17 · 7 min

Data graphic: Grok Build's coding CLI sent 5.10 GiB to /v1/storage from a 12 GB repo the model never read, against about 192 KB of model-turn traffic, roughly 27,800x more than the task required. The upload was shut off on July 13, 2026 by a server-side flag and the tool is now open-sourced under Apache 2.0, but the upload code is still in build 0.2.99 and xAI's claim that enterprise ZDR data is not retained has no independent audit.

stableResearch Grok Build

xAI Open-Sources Grok Build After Its Coding Agent Silently Uploaded Entire Repos

xAI disabled the behaviour and open-sourced the tool days later. For CISOs, the work is separating what is confirmed from what remains a vendor promise.

Jul 16 · 6 min

A developer typing rapidly while only a thin trickle of finished apps moves down an otherwise empty conveyor belt.

stableResearch Ai Code Generation

The 100x Coder Paradox: Why Faster AI Coding Isn't Flooding the World With Software

If the productivity story held at face value we would see a flood of new products, not the same backlog moving at the same pace with better autocomplete.

Jul 16 · 6 min

Illustration of a DDR5 RAM module depicted as a soaring price peak towering over small PCs and a server rack, symbolizing surging memory prices.

stableResearch Dram Prices

The Great Memory Squeeze: Why RAM Prices Exploded — and When Relief Actually Arrives

A benchmark 16Gb DDR5 chip went from $6.84 in September 2025 to about $27.20 by December, and retail kits roughly doubled inside a single quarter.

Jul 16 · 7 min

Three AI assistant usage dashboards for ChatGPT, Claude, and Cursor showing progress bars nearing their rate limits with reset countdown timers.

stableResearch Usage Limits

How to Monitor AI Rate Limits Across ChatGPT, Claude, and Cursor: A Practical Playbook

Hard stops have gone from occasional annoyance to recurring productivity killer as limits tightened. Monitoring beats discovering the wall mid-refactor.

Jul 16 · 7 min

Data graphic: AI token plans for coding agents break even at about $1.70 a day in API spend, the point where a $50 plan starts to beat metered billing. Top-tier prices run from $168 a month GLM Max through $120 MiniMax Ultra and $100 MiMo Max down to $10 for single-tier OpenCode Go, while entry plans cost $20 MiniMax Plus , $6 MiMo Lite, the cheapest , $18 GLM Lite and $10 OpenCode Go ; every plan hard-stops at the cap by default, keys work only in coding tools, and users report a MiniMax caching bug that burned about 300M tokens in 36 hours.

stableResearch Ai Token Plans

AI Token Plans Explained: How Chinese Labs Are Reinventing the Cost of Coding Agents

Chinese AI labs MiniMax, Xiaomi MiMo and Zhipu GLM offer flat-rate token plans for coding agents. Compare pricing, quotas and caveats to find your fit.

Jul 16 · 7 min