In partnership with
Welcome, AI enthusiasts
Fast month, OpenAI's best model launched to just a small group of approved partners. Tomorrow, it opens to everyone. GPT-5.6 is already rolling out worldwide in preview, and the full public launch is a day away. Access went from locked down to wide open in a matter of weeks. Let's dive in!
In today’s insights:
GPT 5.6 Goes Live for Everyone Tomorrow
Anthropic Extends Fable 5 and Shares a Blueprint to Run It for Half
Claude Cowork Now Finishes Tasks While You're Offline
Read time: 5 minutes
LATEST DEVELOPMENTS
Evolving AI:OpenAI is ending the limited preview that held GPT-5.6 back from the public.
Key Points:
GPT-5.6 first launched to only a small group of vetted partners, and now opens to everyone as global preview access widens.
The release spans three tiers, from the flagship Sol down to a cheaper Terra and a low-cost Luna built for speed.
OpenAI calls it its strongest model yet, with its sharpest gains in coding and cybersecurity, shipped behind its most robust safety stack to date.
Details:
GPT-5.6 is rolling out worldwide now ahead of tomorrow's public launch. It brings a new max setting that gives the model more time on the hardest problems, plus an ultra mode that splits complex work across subagents. Sol leads the lineup, while the cheaper Terra matches GPT-5.5 at half the price and the low-cost Luna undercuts both. The family starts at $1 per million tokens, and Sol reaches Cerebras chips this month at up to 750 tokens a second.
Why It Matters:
Sam Altman is putting that cheaper, faster access in front of everyone tomorrow. Inside the same app they already open, GPT-5.6 can take on heavier work, from writing and research to code most people could never write themselves. For many, the low price is what makes the strongest model worth reaching for day to day.
TOGETHER WITH DATADOG
🔐 LLM Observability Best Practices
Evolving AI: 4 Key Insights for Scaling LLM Applications.
LLM workflows can be complex, opaque, and difficult to secure. Get the latest ebook from Datadog for practical strategies to monitor, troubleshoot, and protect your LLM applications in production. You’ll get key insights into how to overcome the challenges of deploying LLMs securely and at scale, from debugging multi-step workflows to detecting prompt injection attacks.
Evolving AI: Anthropic extends more free days for its most capable model, Fable 5, plus a playbook to run it for up to half the cost.
Key Points:
Fable 5 stays included on paid plans at no extra cost for a few more days after Anthropic moved the cutoff.
Alongside it, Anthropic laid out how a cheaper Sonnet 5 model can carry most of a task and call Fable 5 only when the strategy is hard.
The same setup also runs in reverse, with Fable 5 planning the work and handing execution down to cheaper workers.
Details:
Anthropic's new orchestrator pattern reaches 96% of Fable 5's results for 46% of the cost, a little over half off. The lighter advisor setup, which consults Fable 5 only about once per task, still holds 92% of its coding score at 63% of the cost, and each sub-agent keeps its own cache so repeated calls don't pay twice. Anthropic's extension keeps included Fable 5 access on paid plans running through July 12.
Anthropic's plan-big, execute-small cookbook, the working blueprint, step by step.
Why It Matters:
Sonnet 5 doing the heavy lifting means an individual can hand off the bulk of their work to a cheap model and still get Fable 5's judgment on the calls that decide the outcome. The reader who learns this pattern spends far less for near-identical results, and keeps that edge long after the free window closes. Knowing when a task truly needs the strongest model is quietly becoming the real skill.
AlphaSignal's 8-person team automated 18 workflows. No developer.
AlphaSignal runs the most-read technical AI newsletter with an 8-person team. One salesperson installed Viktor for prospect research. Sixty-seven days later it runs 18 workflows across sales, ops, editorial, and finance: proposal builder, deal updates, competitor monitoring, P&L analytics. No developer hired.
CLAUDE COWORK
📱 Claude Cowork Now Finishes Tasks While You're Offline
Evolving AI: Claude Cowork moves to web and mobile and can now complete scheduled work with every device offline.
Key Points:
Scheduled tasks now run with no device online, building briefs from email threads and transcripts overnight.
Cowork arrives on web and mobile in beta, with Max users getting access first over the coming weeks.
Chat and Cowork now share one home on web and desktop, with projects and artifacts in one place.
Details:
Anthropic is bringing Claude Cowork to web and mobile, so a session started at a desk can be finished from a phone. A 6am client prep works through email threads and transcripts, then leaves a briefing doc built and the follow-up email drafted but unsent. That everyday knowledge work makes up more than 90 percent of Cowork use, led by business operations and content creation. Beta reaches Max users first, with doubled usage limits through August 5.
Why It Matters:
Claude Cowork now absorbs the overhead work that fills much of a week yet sits in no job description, like turning a folder of contracts into a risk-flagged tracker. That load is done by morning without a screen in the loop, and the agent still pauses at any call that needs a person. Delegating is quietly becoming how people work with AI.
👀 Our tip
AI agents are moving beyond one-off prompts. This practical guide from Anthropic explains why the future is designing AI systems that can plan, verify, and improve their own work in loops: a shift that's just as relevant for marketers, operators, and founders as it is for developers.
Source: @ClaudeDevs on X
👀 Click on the image you think is real
QUICK HITS
🏭 Anthropic signed a 20-year, $19B TeraWulf lease for a 401MW Kentucky data center campus.
🔐 Sysdig disclosed JADEPUFFER, the first fully AI-agent-driven ransomware attack, encrypting a database autonomously.
🇪🇺 EU Commission presented an AI Action Plan addressing cybersecurity risks from advanced AI models.
🎮 Meta Quietly Launches Pocket, an AI App for Making and Sharing Mini-Games From Text Prompts.
📈 Trending AI Tools
📝 Granola - AI notetaker that captures the real insights and turns every conversation into ready-to-share, action-driving notes*.
🤖 Auto Learning Agents - Open-source self-hosted AI agents that learn from every task.
🛋️ MeltFlex - Redesign any room with real shoppable furniture in seconds.
🌀 Depthflow - Turn flat 2D images into 3D parallax motion.
*partner link