{{first_name | Leader}}, welcome back.
The next AI race may not be about intelligence. It may be about trust. As models start doing more work on our behalf, knowing when they are uncertain becomes a feature, not a flaw.
These are today’s updates.
Explore how unchecked AI permissions create hidden security risks.*
Anthropic launches Opus 4.8
Tools, resources, and what’s trending on socials
A prompt to turn a plan into a weekly tracking system.
. Together with Opal Security.
Every service account, every API key, every AI workflow accumulates permissions that no one revokes. Permissions nobody audits are the new attack path.
Learn why authorization is the quiet breach vector.
. Top News.
Every major AI lab dropped a new model this year. Most of the announcements followed the same script: better benchmarks, faster responses, lower costs.
Anthropic took a different angle with Claude Opus 4.8.
Anthropic says Opus 4.8 is roughly four times less likely than Opus 4.7 to approve bad code silently. It admits uncertainty, flags flaws, and pushes back on unsupported assumptions. Early users report it asks better questions, catches its own mistakes, and challenges bad plans rather than executing them.
That's a deliberate product decision. And it signals where Anthropic thinks the real problem lies.
Under the hood, Opus 4.8 improves coding, browser-based tasks, reasoning, and long-context work. Tool usage is more efficient, context retention is stronger, and API developers can now update system instructions mid-task without breaking workflows. Pricing stays the same as 4.7.
As AI moves from answering questions to actually doing work, silently wrong becomes a much more expensive failure mode than visibly uncertain. A model that says "I'm not sure this is right" is more useful inside a real workflow than one that confidently produces something broken.
. Signals.
Tools
Links
Introducing Pitch Agent. Create a custom deck from a prompt and optional file context, then refine slides via chat — all in Pitch. Try it today.*
. Poll.
What would make you double down on AI investment?
. Market.
Funding
Cypher AI raised $2M (Seed) to build AI-native infrastructure for life sciences R&D.
RevEng AI raised $15M to verify the security and integrity of AI-generated software.
Roles In AI
Senior Technical Recruiter, AI/ML Research at Together AI
Staff Product Manager, Conversion ML at Liftoff
Socials
. Prompt of the Day.
Turn This Plan Into a Weekly Tracking System
When to use this?
Once an action plan exists and you want to ensure it actually moves week to week.
Convert the plan below into a simple weekly tracking system.
Requirements:
Identify what needs to be tracked weekly vs what doesn’t
Define 1–2 signals per action that show progress or risk
Suggest a weekly check-in format (questions to ask, not a meeting agenda)
Flag when leadership should step in
Output sections:
Weekly metrics to track
Weekly questions to ask
Escalation triggers
Keep it minimal and execution-focused.
Plan:
[paste the action plan]P.S. Get more such prompts in the Prompting Playbook (free for you)
Stay curious, {{first_name | leaders}}
PS. If you missed yesterday’s issue, you can find it here.