Good Morning! Here's what I have for you in today's newsletter:
Anthropic introduces Claude Fable 5.2 and Mythos 5.2
Meta’s Muse Ember targets everyday agent tasks at lower cost
Google gives Antigravity a layout for research and writing
Get job-ready in 1 week with Claude Code
4 new AI tools worth trying today
AI MODELS
Anthropic introduced Claude Fable 5.2 and Mythos 5.2 for demanding coding, reasoning, and visual tasks. The company reports up to 30% faster output than the previous versions, while testers noted clearer communication.
On FrontierCode 2.0, Fable 5.2 at High effort roughly matched Opus 5.5 at Max effort for about half the cost per task. At Xhigh, it led the released models tested.
A group of 1,000 Fable 5.2 agents formalized Fermat’s Last Theorem in 30 hours, producing a computer-checkable proof. Under the same setup, Opus 5.5 completed 26% in 18 hours.
Mythos 5.2 developed a new approach to predicting whole capsid structures, the protein shells used in gene therapy delivery, modeling 28 of 34 attempted. Medium effort is the default in Claude apps and Claude Code.
A model’s best score matters less if reaching it makes every task too expensive. Fable’s reported results suggest stronger coding performance without always using the highest effort setting. The math and capsid projects show a different ambition: keeping agents productive through long technical jobs. Those demonstrations are substantial, though they don’t establish how either model will handle every project.
AI AGENTS
Meta introduced Muse Ember, its smallest model, for routine agent tasks. API pricing is $0.10 per million input tokens and $0.50 per million output tokens. It’s also available in Muse Code across every plan and pay-as-you-go access.
Ember can work across apps, make successive tool calls, run tests, and apply targeted code fixes. Meta also says it can operate desktop software using screenshots.
Meta reports performance comparable to Claude Haiku 5.5 on the Artificial Analysis Intelligence Index at roughly one-third of the cost. Cached input is $0.004 per million tokens.
A 1 million token context window supports text, images, audio, and video. Meta suggests using Spark or Wonder for planning and delegating higher-volume work to Ember.
An agent’s bill can grow through dozens of ordinary steps rather than one difficult answer. Ember targets that repeated work, potentially making it cheaper to reserve larger models for harder decisions. Its mixed-input support also covers files, screenshots, and recordings, though teams will need to test whether the savings hold across their actual tasks.
AI TOOLS
Google launched a Knowledge Work layout in Antigravity for research, writing, and generated artifacts. The interface changes, but the agent retains access to the same models, local files, plugins, skills, custom agents, and MCP servers.
Replies, artifacts, and Folders remain visible, while terminals, file diffs, tool calls, and other development panels move aside. A progress line appears during work and clears afterward.
Permission requests now include plain-language explanations. New users can choose roles such as Knowledge Work, Data & Analytics, Science, or Design to set initial layouts, prompts, and plugin recommendations.
Change layouts under Settings > General > Interface Layout without restarting. Google says tens of thousands of employees outside software engineering already use Antigravity weekly.
A terminal panel isn’t much help when you’re reviewing a report. This layout gives documents and responses more room without removing the underlying tools. Clearer permission summaries could also make it easier to understand what the agent wants to do. Users can switch layouts as the work changes rather than move to another app.
HOW TO AI
Check your fit, strengthen your CV, and prepare your applications.
career-ops is a free, open-source tool inside Claude Code that compares job ads with your CV and scores your fit from 1 to 5. You’ll need Claude Code with a paid Claude plan, Node.js, and your current CV.
Step 1: Install career-ops
In Terminal, run npx @santifer/career-ops init, followed by cd career-ops, npx playwright install chromium, and npm run doctor.
Check that the billing result reports no API key in the environment if you intend to use your subscription. Four missing setup files are expected before profile setup. Start Claude Code with claude, then enter /career-ops to view its modes.
Step 2: Teach it your CV
Complete setup through chat. Share your CV, target roles, location, minimum salary, deal-breakers, and work rights. Ask for strict scoring based only on experience you can support.
Request a five-line summary of your preferences and correct anything missing before reviewing jobs.
Step 3: Score five live jobs
Send five current job ads individually using /career-ops followed by each link. The README recommends a 4.0 application threshold.
Alvaro’s test runs took 7 to 10 minutes each. Use /career-ops triage for an initial screen, or paste the ad text if a link fails.
Request one comparison table showing company, role, score, main gap, and recommendation.
Step 4: Fix the gaps and apply
Ask which gaps recur across reports and whether they need clearer wording, existing evidence, or new experience.
Add accurate supporting examples and rescore your strongest role. In Alvaro’s demo test, two additional examples raised one score from 3.8 to 4.2.
Generate the tailored documents with /career-ops pdf and /career-ops cover. Review every claim, submit the application yourself, and record it through /career-ops tracker. Scores guide your review; they don’t predict an offer.
P.S. Become an AI Operator here.
OpenAI added support for connecting several local computers to one dot. Users choose which machine handles each task, such as testing an iOS change in Xcode on a second Mac while continuing to code on a laptop.
ElevenLabs introduced ElevenAgents Campaign Manager for outbound campaigns. Teams supply contacts and an objective, such as qualifying a lead or confirming an appointment. The agent follows up over time and retains context between interactions.
Google brought Maps Platform into AI Studio with one-click setup. Gemini can build maps apps with places, weather, routing, 3D buildings, and Street View. API access doesn’t require a credit card.
🖼️ Seedream 5.5: ByteDance says its image model preserves untouched areas across 100 editing rounds. It also adds native transparent output and sharper small text.
🎬 HeyGen: Launched Platform and Playground for testing its video, voice, and avatar models before using them through the API.
📱 York AI: an AI social media manager that learns your voice from your past posts, then writes, schedules, and posts for you, sending approvals and daily reports to your messaging app.
📽️ Vidsembly: Converts PowerPoint and PDF decks into narrated MP4s, with AI voiceovers, voice cloning, and slide translation for other languages.
THAT’S IT FOR TODAY
Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!
See you tomorrow :)
- Dr. Alvaro Cintas
✓ Full archive of premium guides with ready-to-use prompts
✓ Structured AI courses (step-by-step, start-to-finish)
✓ Every upcoming premium tutorial
✓ $500+ worth of AI tools