Latent.Space
substack
02 May
~60 min
Latent.Space announces Wave 2 call for speakers at AI Engineer World's Fair 2026, emphasizing newly prioritized research and engineering tracks (autoresearch, memory, world models, tokenmaxxing, agentic commerce, vertical AI) alongside comprehensive coverage of the frontier model and infrastructure landscape as of April-May 2026.
Novel signal on agent runtime becoming primary competitive moat over model quality; spatial grounding as explicit reasoning primitive (DeepSeek Vision); latent-space agent communication as efficiency alternative to text—these represent architectural shifts beyond typical model scaling narratives.
surprising_data
debate
The Oracle by Polymarket
· Polymarket
substack
01 May
~9 min
Author argues that the Callais Supreme Court decision narrowing the Voting Rights Act will trigger a redistricting arms race benefiting Republicans by 4 seats in 2026 and up to 12 in 2028, but current prediction markets are underweighting this shift due to competing macroeconomic headwinds.
The quantification of seat-level impacts (4 in 2026, up to 12 in 2028) and the emphasis on state trifectas as the truly consequential outcome beyond congressional redistricting offers granular forecasting detail; however, the core Callais analysis and its electoral implications are already widely discussed in political forecasting circles.
prediction
deep_analysis
surprising_data
Don't Worry About the Vase
· Zvi Mowshowitz
substack
01 May
~32 min
Author argues that housing scarcity is almost entirely caused by regulatory barriers—zoning, building codes, preservation rules, permitting delays, and labor requirements—that make construction economically infeasible, and that most proposed 'solutions' (affordable housing mandates, prevailing wage rules) paradoxically worsen the problem by adding costs and complexity.
While housing regulation is well-covered, the piece's systematic documentation of specific quantified regulatory costs (294x benefit ratio, 20% prevailing wage premiums, 70% rejection rates) and mechanical analysis of how thresholds create perverse incentives (99-unit clustering) provides precision beyond typical YIMBY discourse; the SRO coordination problem framing and Federal preemption solution are underexplored in mainstream coverage.
contrarian
surprising_data
deep_analysis
Author argues that our understanding of Michael Jackson is fundamentally fragmented by generational exposure—different age cohorts absorbed entirely different narratives about him (child prodigy, pop icon, tabloid freak, or accused criminal) depending on which era's cultural output they encountered, making unified judgment impossible.
Dan Davies - "Back of Mind"
Author argues that accountability sinks are not merely intentional managerial evasion tactics but rather inevitable consequences of industrialized decision-making systems where causality becomes genuinely multi-layered and ordinary language concepts of blame cannot map cleanly onto complex institutional contexts.
Author argues that a assassination attempt at the White House Correspondents' Dinner reveals a dangerous normalization of political violence on the left, evidenced by widespread online celebration of the attack and deep partisan distrust (47% of Democrats believe the July 2024 attempt on Trump was faked).
DeLong's Grasping Reality: Economy in the 2000s & Before
Author argues that human cognitive advancement from -70,000 to -3,000 occurred at less than 0.007% annually because early societies, despite possessing Anthology General Intelligence (AGI), devoted minimal effort to recursive technological self-improvement and accumulated mostly local, non-transferable knowledge.
22 more not shown