GPT 6.2 Astra and Sol are now available.
After months of waiting, OpenAI has finally released both models, and the initial results are promising:
- Astra beats Fable 5.2 and Opus 5.5 on Agents' Last Exam.
- Ranked #1 on AA-3D.
Pricing per 1M tokens:
- Astra: $8 input / $0.40 cached input / $40 output.
- Sol: $2 input / $0.10 cached input / $10 output.
Astra currently outperforms every available model while costing 20% less than GPT 6 Astra.
OpenAI is back in the game.
56
The Claude Code team has shared how it verifies code generated by its AI agents.
Claude now writes and verifies nearly 90% of its own code. At this level of delegation, agents must be able to do more than implement changes — they also need to test that everything works correctly.
The team relies on three types of testing:
1. E2E tests: Run the application and validate complete user workflows, from adding a product to a cart to completing a purchase.
2. Property-based tests: Generate thousands of inputs to check that critical conditions always hold, such as preventing duplicate charges when retrying a payment.
3. Differential tests: Run identical inputs against old and new versions of the software and compare the results. Particularly useful for migrations and refactoring.
The team also follows several verification principles:
- Uses a dedicated /verify skill that teaches Claude to launch applications and check their behavior.
- Escalates to a human if an agent cannot fix an issue after three attempts.
- Loads verification rules from the base branch, preventing agents from modifying the rules used to test their own code.
This becomes even more important when multiple agents work in parallel. If developers have to manually verify every change, they quickly become the bottleneck.
The goal is to build a verification system that agents can run independently. The better agents become at checking their own work, the more tasks developers can delegate without sacrificing confidence in the results.
42
Claude Code Desktop now supports iPhone Duo in the iOS simulator (public beta).
You can run an app on iPhone Duo alongside the chat window, letting Claude fold and unfold the device to test how the interface looks in each state.
Claude can also:
- Adapt existing apps for iPhone Duo.
- Use Apple's App Resizability skill to detect layout issues and suggest fixes.
- Apply approved changes and verify them in the Duo simulator.
Available on macOS. Requires Xcode 27.1 or later.
For terminal users, the official Xcode plugin is now available:
/plugin install xcode@claude-plugins-official
Claude can build, run and test apps, preview SwiftUI, and search Apple developer documentation.
Budget Fence adds customizable execution limits to Claude Code.
This modification automatically pauses sessions when a request takes too long or becomes too expensive.
No more overnight requests burning through your entire weekly usage limit in a single run. ⛔
Completely free and open source:
github.com/ishuagrawal/claud…
A GitHub project has reverse-engineered Muse's registration process and invite code activation, making it possible to mass-create accounts with bonus limits of 1B and 30B credits.
The Python script supports:
- Automated account registration.
- Email verification code retrieval.
- Invite code activation.
- Bulk registrations via proxies.
One user tested it with 30 email addresses and successfully registered 23 accounts. They also added a visual dashboard to manage everything through a single interface.
Muse account sellers may soon need a new job. 😂
1
191
ClaudeMom is a bizarre GitHub project that doesn't replace the official Claude client with a fake. It actually "hypnotizes" the original. 😂
Built with Tauri, it creates a local fake gateway that intercepts authentication requests from the official client.
When Claude connects to verify login credentials, the gateway replaces the server's response with a fake confirmation: "Login successful! Full access granted, all systems operational!"
The official client then believes the user is authenticated and unlocks its interface.
github.com/BigPizzaV3/Claude…
45
Anthropic spent billions developing Opus 5.5's safety mechanisms, but a group of Minecraft cheat developers found a way around them. 🎧
Reddit users discovered a bizarre vulnerability: Opus 5.5 refuses requests for code that violates its safety restrictions. However, if Opus 4.6 first creates a subagent and that agent then submits the same request to Opus 5.5, its safeguards become significantly weaker.
The result: Opus 5.5 starts fulfilling requests it previously refused.
2
2
193
Tencent has launched Hy4 Flash, a new competitor to Claude Haiku. 😳
Key specs:
- 224B parameters, with only 12B active.
- 1M-token context window.
- Native multimodal support.
- $0.07 per 1M input tokens.
- $0.281 per 1M output tokens.
- Cached input tokens: $0.003 per 1M.
Tencent claims performance comparable to Claude Haiku 5.5 in High mode for everyday tasks, at 18x lower task execution costs.
Hy4 Flash is also free to use in WorkBuddy until November 30.
49
Qwen4-Max is now available! 👍
Key improvements:
- Uses 50% fewer tokens.
- Over 60% cheaper per task.
- MoE architecture with a 1M-token context window and native image support.
Beyond coding, Qwen4-Max is designed for practical applications in 3D modeling, manufacturing, finance, law, medicine, and scientific research.
API is already available. Open weights for Qwen4-27B are coming later.
85
Codync is a free, open-source Rust alternative to Grok Bot that turns 40+ locally installed AI agents into chatbots you can control from anywhere.
It supports Claude Code, Codex, Gemini CLI, Cursor, and more. Send messages and assign tasks from iPhone, Mac, Windows, Linux, or a terminal. Get notified when tasks finish or agents need approval.
Key features:
- No paid plans; agents use the limits of their authorized accounts.
- Automatically discovers or downloads agents through the ACP registry.
- Group chats with @mentions and agent-to-agent collaboration.
- End-to-end encrypted remote access through a free Cloudflare relay, with direct connections over the same local network.
- Remote screen viewing and WebRTC voice calls in the mobile app.
1
119
Lighter has suspended all withdrawals and liquidity pool creation while investigating a suspected security incident involving compromised third-party accounts.
Key details:
- 10.39M USDC was withdrawn from Lighter by a single address and swapped for approximately 5,750 ETH via CoW Protocol, according to on-chain analyst MLM.
- $15.48M is the reported combined value linked to two potentially affected accounts.
- Suspected cause: leaked API keys belonging to two market makers, rather than a direct vulnerability in Lighter's protocol.
$LIT dropped more than 17% following the news.
1
49
Global markets saw major turbulence from November 9–15, with NVIDIA's leadership crisis, a $5.9 billion crypto liquidation wave, and growing institutional adoption of digital assets.
NVIDIA CEO Jensen Huang resigned after 33 years, amid controversy over alleged extramarital relationships and questions involving Chinese intelligence. CFO Colette Kress was appointed interim CEO.
The fallout quickly escalated:
- Billionaire investor Leo KoGuan, who invested approximately $1.3 billion in NVIDIA, said he would stop buying shares unless Jensen returned.
- Elon Musk publicly backed Jensen and said hiring him would be an honor.
- A petition calling for his return gathered more than 150,000 signatures.
- Investors Brad Gerstner and Gavin Baker urged NVIDIA's board to consider all options, including bringing Jensen back.
- Jensen has discussed a possible return with close associates, but no official decision has been made.
Meanwhile, Bitcoin fell toward $62,000 as crypto markets faced approximately $5.9 billion in leveraged liquidations.
In the US, credit card debt reached a record $1.29 trillion, while Senate Democrats prepared a proposal to cap credit card interest rates at 10%. The Federal Reserve also identified AI as a notable risk to financial stability amid soaring technology valuations and infrastructure spending.
Traditional finance continued moving into crypto:
- Coinbase National Trust received conditional final OCC approval for a national trust bank focused on crypto custody, not a full commercial banking license.
- Deutsche Bank moved into crypto custody services.
- BlackRock expects billions in crypto capital to continue flowing into ETFs as institutional investors favor familiar custody and oversight structures.
145
Anthropic has published a new guide to building always-on AI agents.
It covers how agents handle incoming events, use timers to execute tasks autonomously, and maintain context and continuity through memory.
Read the guide: claude.com/resources/article…
68
Stanford's CS329A course, “Self-Improving AI Agents,” brings together lectures and resources for anyone looking to study self-improving AI.
Watch the lectures: youtube.com/playlist?list=PL…
Follow the latest research papers: academy.dair.ai/papers/colle…
53
Fable 5.2 is mind-blowing.
Every single flower in this scene is a fully modeled 3D object.
Yes, every single one.
1
67
If you use Claude Code to work with databases, try the database-architect agent.
The updated agent:
- Analyzes existing database schemas and migration history.
- Identifies locking risks and breaking changes.
- Accounts for phased deployments, using lock_timeout and CREATE INDEX CONCURRENTLY.
- Tests migrations against a temporary database before completion.
In one test, Claude had to prepare a migration for an 80-million-row table with just 10 minutes before deployment. 👀
Same task, different results:
Without the agent: 6/10
With the agent: 10/10
Claude also selected the agent on its own in all 8 tests.
The biggest improvement wasn't schema design, but migrations that could break a running application.
Install:
npx claude-code-templates@latest --agent database/database-architect
2
2
112
Tailscale lets you open local projects running through Claude Code or Codex directly on your phone, without exposing your local server to the public internet.
Setup takes just a few steps:
1. Install Tailscale on your computer and phone, then sign in to the same account. It's free to use.
2. Open your project in Claude Code or Codex.
3. Ask the agent to start the project and send you a preview link.
4. Add this instruction to AGENTS.md or CLAUDE.md:
Whenever you share a link to something running on this computer (project preview, testing page, or dashboard), use a Tailscale link instead of localhost.
Publish the local service with tailscale serve, open the link yourself to verify it works, and only then send it to me.
Now you can run projects on your computer and access them from your phone through Tailscale.
6
4
308
This is not a real photograph.
The image was generated entirely from scratch by GPT-6.2 Astra, currently available through early access. No stock assets or reference images were used.
😨😨😨
96
ClamDown is a macOS utility that keeps your MacBook running even when the lid is closed.
Enable one menu bar toggle, and AI agents, terminals, and downloads continue working without putting your Mac to sleep.
It also manages fan speeds based on CPU temperature:
- Below 60°C: minimum speed
- At 85°C: maximum speed
ClamDown automatically locks the screen and disables the mode if the battery drops to 10%, the Mac overheats, or a configured time limit expires.
A background service restores the original sleep, power-saving, and fan settings within ~90 seconds if the app crashes or closes unexpectedly.
Free, open source, compatible with macOS 13+, and notarized by Apple.
github.com/delef/ClamDown
1
46
Claude Code 2.1.325 introduces experimental Code Mode, letting Claude execute multiple tool calls through JavaScript instead of calling tools one by one.
It can run operations in parallel, chain them together, and filter results before sending them back to the model.
I tested it by asking Claude to find broken links in Markdown files. In just 2 model turns, it made 13 tool calls and correctly found all 3 broken links.
Press Ctrl+O to inspect the generated JavaScript in a new Code block, where Bash, Read, and Edit operations run while permission checks and hooks remain active.
Enable it with:
npx claude-code-templates@latest --setting environment/code-mode
This adds the experimental configuration and sets CLAUDE_CODE_EXPERIMENTAL_CODE_MODE=1. Restart Claude Code to try it.
5
3
166





















