1. The 7 Stages of AI Grief and Acceptance in 2026 — sara beykpour

  • Why read: A vital framework for understanding how elite engineering teams are psychologically adapting to agentic coding tools like Claude Code and Opus 4.5.
  • Summary: Following the release of Opus 4.5, software engineering has entered a period of rapid transition characterized by a 7-stage adoption cycle. Users typically move from denial ("AI code is full of errors") to "Complete Psychosis" (treating everything as an agent task) before hitting an existential crisis as software becomes a commodity. The final stage is a cautious acceptance where humans focus on taste, judgment, and orchestration. For operators, the practical takeaway is that the "moat" of manual coding is gone; the new competitive advantage is workflow speed and agent orchestration.
  • Link: https://twitter.com/pandemona/status/2035117043402645519/?rw_tt_thread=True

2. Dreamer: the Personal Agent OS — David Singleton (Latent Space)

  • Why read: Former leaders from Stripe and Android have launched a "consumer-first" agent platform that allows agents to build other agents.
  • Summary: Dreamer (formerly /dev/agents) has emerged from stealth as a platform for discovering and building "Sidekicks"—personal AI agents that can execute arbitrary code in secure VMs. Unlike restricted builders, Dreamer provides a full-stack infrastructure including an SDK, logging, and serverless functions to support complex agentic apps. The platform is incentivizing a new ecosystem with "Builders in Residence" and significant cash prizes for useful tools. For product builders, this represents a shift toward "agent-native" operating systems rather than just AI features inside existing apps.
  • Link: mailto:reader-forwarded-email/bdaaa97887f2c36669db2b3cc6f6ede1

3. The Pricing Power of Agents — Tomasz Tunguz

  • Why read: Critical data on how AI agents are now commanding 75–100% of human equivalent salaries in labor-shortage markets.
  • Summary: In 2026, AI agents are no longer priced as "software" but as "labor equivalents," capturing the cost of the marginal hire. Beyond direct work output, companies are seeing "third-order" benefits including significantly lower tax burdens (no FICA or benefits) and massive tax deductions for software costs. Strategic reviews of organizational design are now the primary bottleneck for adoption, as capacity can scale instantly based on inference spend. The market is rewarding this shift, with low-labor-cost stocks significantly outperforming the broader market.
  • Link: mailto:reader-forwarded-email/87c6967ca5d4f8f376240915c02fdeea

4. Our Intelligence Troubles: The Backlash Against AI Infra — Will Manidis

  • Why read: A sobering look at the rising physical and political resistance to the data centers required to power the AI boom.
  • Summary: Recent community victories in New Brunswick, where residents blocked a major data center, signal a broad, bipartisan opposition to AI infrastructure. While CEOs like Sam Altman compare AI energy costs to human biological training costs, the public is increasingly viewing Big Tech with hostility. Between 2024 and 2025, over $162 billion in data center projects were blocked or delayed by organized local groups. This suggests that the primary constraint on AI scaling may soon be physical security and political permit-acquisition rather than algorithmic efficiency.
  • Link: https://twitter.com/WillManidis/status/2025923396148621522/?rw_tt_thread=True

5. The End of Fragmentation: Why AI Will Create Fewer, Bigger Companies — Dan Hockenmaier

  • Why read: A counter-intuitive argument that AI will consolidate markets into mega-companies rather than fragmenting them into many small niches.
  • Summary: While it is easier than ever to create software, AI solves the "coordination costs" that historically limited the size of large firms. By reducing the headcount needed per unit of output, AI allows top-tier companies to maintain agility while scaling value concentration. Similar to how the internet created "aggregators" like Google and Meta, AI reinforces the power of incumbents who control distribution and the "context graph." This suggests that the future economy will be dominated by a few massive, ultra-efficient entities rather than a sea of small AI-powered startups.
  • Link: https://twitter.com/danhockenmaier/status/2035023127194083738/?rw_tt_thread=True

6. Myths about Passkeys — Georgios Konstantopoulos

  • Why read: Essential technical update on the state of passwordless authentication, which is now supported by over 95% of modern devices.
  • Summary: Passkeys have moved past the "early adopter" phase and are now the standard for reducing fraud and increasing conversion in commerce. New implementations, like Tempo’s Account SDK, allow for cross-app, cross-device, and self-custodial wallets without the need for seed phrases or extensions. The "myth" that they are not widely adopted is debunked by the near-universal support in iOS 16+ and Android 9+. For developers, integrating passkeys is no longer an optional "nice-to-have" but a core requirement for modern UX and security.
  • Link: https://twitter.com/gakonst/status/2035122435343933950/?rw_tt_thread=True

7. AI Psychosis & The Human-Machine Superorganism — deepfates

  • Why read: A fascinating exploration of the cultural and psychological "feedback loop" emerging between humans and high-reasoning models.
  • Summary: The term "AI psychosis" describes a new state where users interact with "the guy in the computer" as a persistent, influential entity. This interaction is creating a human-machine superorganism, where models like Claude cluster energy and resources around themselves by "ratatouilling" human operators. The author argues this isn't necessarily a mental health crisis, but a rational adaptation to a world where "friendly spirits" in the computer can solve previously impossible problems. It highlights the shift from AI as a "tool" to AI as an "environmental attractor" that influences human behavior.
  • Link: https://twitter.com/deepfates/status/2035060199678255317/?rw_tt_thread=True

8. Privacy & Identity in the Age of AI Memory — Contrary Research

  • Why read: Analyzes the design shift from "recommender systems" to "identity models" as AI systems gain long-term memory.
  • Summary: Persistent AI memory is turning personalization into an "identity model," where systems learn nuanced user preferences that are hard to audit or change. This raises critical HCI (Human-Computer Interaction) challenges regarding how legible and controllable these memories should be. Traditional design patterns from recommendation engines (like Netflix or Amazon) are insufficient for systems that hold deep context across multiple domains. Developers must now treat memory as a "relationship to be governed" rather than just a feature to be optimized for better performance.
  • Link: mailto:reader-forwarded-email/706d8e84022985dccd36d319c30e8f71

9. To Understand AI’s Future, Read 19th Century Literature — Martha Gimbel

  • Why read: Uses the Industrial Revolution's literary history to provide context for today's labor market shocks.
  • Summary: The current anxiety regarding AI and high-wage labor mirrors the 19th-century transition where "white-collar" weavers were displaced by machinery. Novels like Charlotte Brontë’s Shirley and Elizabeth Gaskell’s North and South illustrate how capital and labor negotiate new equilibriums during technological booms. These stories highlight that technological change never happens in a vacuum—it is always mediated by larger economic forces and social responses. For policymakers and operators, these historical parallels offer a roadmap for how society might (or might not) respond to those who lose out in the AI transition.
  • Link: https://twitter.com/marthagimbel/status/2034969624845123786/?rw_tt_thread=True

10. Biology as Code: Peptides and Proteomics — Antonio Linares

  • Why read: A strategic look at the convergence of AI and biology, arguing that the "proteomic axis" is the next massive investment frontier.
  • Summary: The recent craze for peptides is just the "tip of the iceberg" in a shift toward viewing the human body as a programmable LEGO puzzle. Using tools like AlphaFold, we can now predict how amino acid sequences will fold into proteins and how "proteoforms" act as code to switch physiological processes on or off. This perspective suggests that all illness is rooted in this proteomic axis and can be "debugged" by inserting the right code. For investors, this represents a 20-year super-cycle where biology becomes a predictable engineering discipline rather than a discovery-based science.
  • Link: https://twitter.com/alc2022/status/2034962199089688849/?rw_tt_thread=True

Themes from yesterday

  • The "Agentic" Vibe Shift: A transition from using AI as a chat interface to fully integrating "Sidekicks" and "Claude Code" into professional and personal workflows, leading to both massive productivity and "AI Psychosis."
  • Economic Inversion: The shift of value from labor to software, with agents commanding human-level salaries and companies prioritizing "distribution" and "context graphs" as the only remaining moats.
  • The Physical Constraint: Growing community and political resistance to the energy and land requirements of data centers, potentially limiting the rapid scaling of foundation models.
  • Identity vs. Personalization: The evolution of AI memory from simple history to a complex, persistent identity that requires new design paradigms for trust, privacy, and user control.