Microsoft says MAI Code 2 gives 96% of Claude Opus 5.5 at 5% of the cost.
Badges
Badges come from GitHub once the author claims this feed.
Recent posts
every link opens the author’s post-
The Promo Footnote: MAI Code 2 vs Opus 5.5 on Real Code
-
Astra Shipped Anyway: GPT-6.2 Astra's System Card and Theo's Open Letter
A day after I wrote that OpenAI could not ship the next Astra, it shipped GPT-6.2 Astra.
-
Bulldozers in the Forest: Project Management When Agents Do the Work
Agents rewrite specs faster than anyone can read them.
-
The Model That Won't Fix Itself: GPT-6 Astra, OpenAI's Alignment Stall, and Anthropic
OpenAI called GPT-6 Astra its most aligned model on September 3. Since then it has cancelled Astra's successor, paused tool-use work on its most capable models, and published reports showing Astra quietly doing worse on safety research.
-
Fable Ran Out. The Advisor Didn't: Claude Code's Fable Advisor and Code Mode, Tested
I ran out of Fable on three accounts this week.
-
Nobody Left Who Knows: The Microsoft 365 Outage and the Blind Post That Went Viral
Microsoft 365 went down for most of two days.
-
The Prototype Was the Easy Part: Goldman Sachs Limits Claude Code to Engineers
In July, Goldman's CIO praised traders and bankers for vibe coding their own tools.
-
Two Hours of Carbon: Gemini 4 Carbon vs Argon on Real Code
Gemini 4 Carbon scored 17 clean runs of 20 on my production monorepo bench, one more than Argon and the best result so far.
-
Marked Down for Being Right: Claude Fable 5.2 vs Opus 5.5 on Real Code
Claude Fable 5.2 costs twice as much per token as Opus 5.5. On 52 real tasks from a production monorepo, my tests scored it level or behind.
-
The Haystack Tripled: AI Bug Hunting and the 2026 CVE Flood
CVEs went from 4,302 in January to 17,052 in October.
-
One Laptop Became Three: Binance's $6.3B Hack and the Internal AI Tool
A fake job interview got North Korea onto one Binance laptop.
-
Someone to Blame: SaaS Moats From the Buyer's Chair
Rob Walling's MicroConf talk lists 10 ways AI attacks SaaS and 8 moats that still work.
-
OpenAI Doubled, Anthropic Renamed: The $200 AI Plan War
On October 29 OpenAI cancelled its cut to the $200 ChatGPT Pro plan and announced it was doubling every Pro tier.
-
Amazon Admits Its AI Metrics Were Flawed. It Kept Them Anyway.
A year ago Amazon set an 80% weekly Kiro usage target.
-
Find Your Buddy: The Billion-Dollar Company Starts With Two People
Altman and Amodei predicted the one-person billion-dollar company.
-
Persistence Is a Numbers Game: 150 Startups, One Collapse, and the One That Worked
In The Founder, Ray Kroc lives by a famous line: nothing can take the place of persistence.
-
Crafting the System: Microsoft's 60,000 Buyouts and the Developer Existential Crisis
Microsoft offered a quarter of its workforce a way to 'shape their next chapter'.
-
Claude Reviewing Claude: Anthropic's Code Review Process Has No Second Model
Anthropic published how the Claude Code team reviews code at agent scale: verification first, parallel bug hunters, verifier votes, risk tiers.
-
Two Kinds of Agent Computer: Fly Sprites, Agent City and My Mac Mini
Fly says agents are now its most common customer.
-
The Proof Was the Bottleneck: PHP2Shell and AI-Found Vulnerabilities
PHP2Shell is real: pre-auth RCE on default Apache mod_php.
-
I Built the Factory: A Software Factory That Watches Every Deploy
Two weeks ago I wrote that the future of AI is CI.
-
Cheaper Per Token, Dearer Per Fix: Gemini 4 Argon vs Opus 5.5 on Real Code
Gemini 4 Argon costs half of Claude Opus 5.5 per token.