Michael Serrano

[Michael Serrano @m_y_serrano]

What's your take on Local LLMs? Is Qwen 3.6 27B still the best for coding & TPS?

[Michael Serrano @m_y_serrano]

Is it just me or is Claude weekly usage much worse than before? I don't understand where the 50% boost is coming from.

[Michael Serrano @m_y_serrano]

Find us on Launch YC! https://t.co/crMcDvmutI

[Michael Serrano @m_y_serrano]

Ever tried to automate a real website and ended up babysitting Playwright selectors, browser-use agents, or computer-use agents?

Watch me use Rindler to set up my plans for the world cup ⚽️🏆
https://t.co/Ut39U6aVTm

[Michael Serrano @m_y_serrano]

We kept running into the same problem: the browser automation would work for a bit, then drift, break, or get lost.

So, we built Rindler to map websites into actions agents can use more reliably.

[Ray Liao @rayruizhiliao]

Today we're launching Inkbox: the identity and communication layer for AI agents. We give agents their own email, phone, iMessage, and an internet address. Install @inkbox_ai identity with one command: allow anyone to reach your Claude Code, Codex, or Hermes from anywhere, and let it reach anyone.

[Michael Serrano @m_y_serrano]

So who was getting heard 😭

[GitHub @github]

We heard you. And we agree.

In light of recent developments in physical media, GitHub is proud to announce that you can now obtain your public repo on CD-ROM.

Keep it. Lend it to friends. Pass it on to your children.

Your code is physically yours, forever. Until you lose it, let's be real.

Order yours today.
https://t.co/z041pdMH7h

[Michael Serrano @m_y_serrano]

Some takeaways:

1. Background agents no longer stop to ask permission when they finish they commit, push, and open a draft PR on their own. There are also new hooks that notify you when an agent needs input or completes. You can genuinely start a task and leave.
2. Subagents got better too. The Explore agent used to run on haiku regardless of what you were paying for, but now it inherits your session's model.
3. aaand the /agents wizard is gone. You just tell Claude to create or manage subagents, or edit .claude/agents/ yourself.

[Claude Code Changelog @ClaudeCodeLog]

Claude Code 2.1.198 has been released.

32 CLI changes

Highlights:
• Claude in Chrome is generally available, offering direct browser access to sessions and agents with no install
• Background agents auto-commit, push, and open draft PRs in worktree after finishing code, automating delivery
• Docs now advise grep for searches and clarify head_limit/unlimited/offset semantics to reduce surprises

Complete details in thread ↓

[Saurabh Suri @surim0n]

how to maximize your fable 5 usage:

1. spend it on the one-way doors. data model, API contracts, the core abstraction - the calls that are expensive to reverse once your week ends

2. put GPT-5.5 in front as a context compressor. It reads the whole repo and hands Fable only the distilled data. Input tokens are precious

3. let fable do your planning. Fable writes the architecture/PRD. you can use the codex plugin inside cc to implement (https://t.co/zBIfEpYZWD)

4. have Fable write the artifacts that govern your cheap fleet - eval suites, grading rubrics, the system prompts GPT-5.5 runs against all week. One frontier session shapes thousands of downstream calls, so the cost amortizes to near zero per use.

REMEMBER: you need to reduce your opus 4.8 usage as well

[Michael Serrano @m_y_serrano]

@mattshumer_ Half usage until July 7 and then API billing afterwards, hopefully this works as an agent orchestrator because small tasks ain't it

[Michael Serrano @m_y_serrano]

@_arthurdls_ Capital One Travel

[Michael Serrano @m_y_serrano]

Can't wait to use my super frontier model and not be allowed to code

[Claude @claudeai]

Fable 5 is back.

[Michael Serrano @m_y_serrano]

@NousResearch Has anyone tried the large page saving feature?

[Michael Serrano @m_y_serrano]

The last thing Fable needed was even more coding guardrails... god forbid it tries to write Python

[Anthropic @AnthropicAI]

Claude Fable 5 will be available again globally tomorrow.

After a series of productive conversations with the US government, we're redeploying the model with a new set of classifiers to target and block more cybersecurity tasks. In the near term, some routine tasks like coding and debugging will fall back to Opus 4.8. We’ll continue to refine these classifiers over the coming weeks to reduce false positives and better distinguish genuine misuse from legitimate requests.

We’ve also begun drafting a consensus framework—with Amazon, Microsoft, Google, and other Glasswing partners—for assessing the severity of AI jailbreaks and how AI developers should respond to them. We invite other industry partners and model providers to join us in this effort.

Finally, we’re scaling up our collaboration with the US government on model testing and safeguards. This will include pre-release access to models and safeguards for evaluation, information sharing on jailbreaks and misuse, and dedicated resources for joint research.

Thank you to our users for your patience, and to our partners across the government, industry, and the research community who worked alongside us to make Fable 5 available again.

Read our full blog: https://t.co/VHyum831ri

[Michael Serrano @m_y_serrano]

What is the first thing that you're going to use Fable for when it comes back?

[leo 🐾 @synthwavedd]

🚨 BREAKING: Commerce Sec. Howard Lutnick has officially lifted the export control order on Anthropic's Fable 5