“Technology is the application of knowledge to control the natural world. It’s the greatest driver of both human prosperity and our capacity for self-annihilation.”
― Naval Ravikant
* Today’s newsletter is brought to you by Auth0.
Production-ready identity for AI Agents
If you're building AI agents, you've probably hit the same wall everyone does.
Hard-coded API keys leak into logs.
Agents read data they shouldn't.
Prompt injections expose user credentials.
One wrong tool call costs you $1k.
You spend more time securing your agent than shipping it.
Auth0 fixes that.
Token Vault hides 3rd-party credentials from agents.
FGA stops cross-user data leaks.
Asynchronous Authorization pings a human before high-stakes actions ship.
Build, launch, and scale on the free tier.
Hi friends,
We hand-picked 3 good pieces of content to inspire and motivate you now and in the future. Keep track of the insights that resonate with you by highlighting them with Glasp💡
If you want to reread or highlight this newsletter, save it to Glasp.
📚 3 Good Recommendations
Securing AI agents: the defining cybersecurity challenge of 2026
by Amit Karp, Mike Droesch, Yael Schiff, and Elliott Robinson (9 mins)
AI agents are now the #1 cybersecurity threat: 48% of security pros see agentic AI as the most dangerous attack vector, with shadow AI breaches costing $4.63M on average. Compromised agents can escalate access faster than humans can respond.
Security requires three stages: Visibility (inventory all agents and permissions), configuration (continuous posture management to limit blast radius), and runtime protection (real-time detection and intervention at machine speed).
Treat agents as identities, not tools: Each agent needs scoped credentials, action-level guardrails, and auditing like any employee—start narrow, expand deliberately.
Getting AI Agents into Production, Fast
by David Brunner (11 mins)
Speed is the surprise — agentic AI workflows can run in 2-3 weeks versus months for traditional software, because you describe tasks in natural language rather than coding every rule.
Hold four lenses at once — decomposition, context engineering, explainability, and evals must be worked simultaneously, not sequentially; they inform each other through iteration.
Ship rough, then refine — get something running in days, study failures to diagnose what’s broken, and iterate fast; the bottleneck isn’t model intelligence but the system around it.
Taste Is the New Bottleneck: Design, Strategy, and Judgment in the Age of Agents and Vibe-Coding
by Itamar Medeiros (21 mins)
Taste is the new bottleneck: When AI handles execution, human value shifts to judgment—discriminating among abundant viable options, not generating them.
Taste is learnable: It’s not innate “eye” but trained judgment—developed through exposure, comparison, critique, and reflection. Teams can build it collectively.
Taste operates at three levels: Deciding what problems are worth solving, judging what’s “good enough” among AI-generated options, and governing what standards get encoded into systems at scale.
📣 Community Updates
🟨 Pricing Update:
On May 1, 2026, Glasp will update its pricing for the first time since launch. The Pro Plan will move from $12 to $15/mo, and the Unlimited Plan from $30 to $36/mo. Existing subscribers won’t be affected until their next billing cycle. The Free plan remains unchanged. See full details here.
🟦 Top 20 Kindle Books of Q1 2026:
We just published our quarterly roundup of the 20 most-imported Kindle books on Glasp — with the single passage that the most readers independently highlighted from each title. See the full list here.
🟩 Post Milestone, 7,000:
We’re thrilled to announce that the number of posts published on Glasp exceeded over 7,000! 🎉 We’ve seen so many amazing insights shared by our community. Thank you for writing and contributing such great content. If you haven’t tried it yet, give it a try here.🟥 Auto-Sync Kindle Highlights:
We published a new tutorial on Auto-Sync Kindle Highlights — a new (Beta) feature that automatically imports your Kindle highlights to Glasp on a daily, every-3-days, or weekly schedule. No more manual syncing. Check the tutorial.
❤️ Gratitude
Thank you for sharing and mentioning us on X, LinkedIn, and/or your blogs. 🙂 We appreciate your support! Please don’t hesitate to ask us anything at any time. Also, feel free to join our Reddit Community ;)
Glasp: Audio Transcriber on FactCheckTool.com
Glasp on i10X
Digital Tools That Can Support Government Social Media Managers on Social Media Strategies Summit
Glasp on TopTools.ai
Glasp for reading later on curtismchale.ca
Chrome Just Got Smarter: 11 AI Extensions That Save You Hours Every Week 🚀 on AI & Tech Insights
AI Browser Extensions: Pros/Cons and 8 Extensions to Know in 2026 on Seraphic
Glasp Review: Features, Use Cases & Alternatives on Future Directory
Islamic Soft Power in a Contested World Cup: YouTube Media Framing and Audience Responses to Qatar 2022 on Arab Media & Society
We hope you enjoyed reading this newsletter!
See you next week ;)
Best,
Kei and Kazuki
--
Would you like to take Glasp on the go?
With the Glasp mobile app, you can highlight and organize your favorite content anytime, anywhere. Stay productive on the move and never miss an insightful quote.
Partner With Glasp
We currently offer newsletter sponsorships. If you have a product, event, or service you’d like to share with our community of learning enthusiasts, sponsor an edition of our newsletter to reach engaged readers.
For anyone who has actually shipped an agent: what broke first?
I'll go first: permissions. It was trivial to make the agent capable and much harder to make sure it only did what it was allowed to do, on behalf of the right person.
If you are earlier than that and still deciding whether to ship one at all, say so below. That is the more common position than the internet suggests, and it is worth talking about honestly.
I really like how this newsletter connects all the dots around where AI is going. The point about agents being both an incredible productivity boost and a major security risk feels very real right now, and treating them like identities instead of just tools is a smart shift. The “ship something rough, then refine fast” approach to agents also matches how the best teams seem to be working today. And the idea that taste is the new bottleneck really stuck with me — once AI can generate endless options, the real skill is knowing what’s worth doing and what “good” actually looks like.