Sarmadi AI Digest July 19, 2026 Updated 7:00 AM CT Today Archive Topics Saved Subscribe RSS

NYC to require AI disclosure in real-estate listings; OpenAI trims Codex context 372k→272k

NYC is moving to require landlords and realtors to disclose AI use in property listings (488 HN) — the sharpest municipal AI-content-disclosure rule yet, arriving days after San Francisco's nudify-app cease-and-desist. A widely-shared essay 'AI Mania Is Eviscerating Global Decision-Making' (243 HN) went viral against the backdrop of Rimer's 'money coming back out' framing. OpenAI quietly reduced Codex model context from 372k to 272k tokens — a rare tier walkback. Dave Eggers told OpenAI staff at their own event that ChatGPT is 'silencing an entire generation' (The Verge). Simon Willison notes Claude Code now runs Bun-inside-Rust. A quiet Saturday leaning on cultural signal and a Codex regression the enterprise procurement teams will notice.

5 papers 7 news 4 sources ← Latest

News

6 items

NYC AI-disclosure rule; 'AI mania is eviscerating global decision-making'

NYC is moving to require landlords and realtors to disclose AI use in listings (488 HN) — sharpest municipal AI-content-disclosure rule yet, days after SF's nudify order. A widely-shared essay 'AI Mania Is Eviscerating Global Decision-Making' (243 HN) went viral against Rimer's 'money coming back out' framing from yesterday.

News Hacker News

NYC may require landlords and realtors to disclose the use of AI in listings

PetaPixel via HN (488 points): NYC would require landlords and realtors to disclose AI in property listings.

Why it matters
  • Sharpest municipal AI-content-disclosure rule yet — applies to a $trillion-scale housing market.
  • Direct sibling to the SF nudify-app order and Google's AI-ad-disclosure policy.
  • Materially affects real-estate photo pipelines and the AI-image-generation tools serving them.
  • Sets the reference template other cities will consider for consumer-facing AI disclosure.
policyregulationsafety
Source →
News Hacker News

AI Mania Is Eviscerating Global Decision-Making

Widely-shared essay (243 HN points) arguing AI mania is degrading real-world decision quality at scale.

Why it matters
  • First widely-cited long-form 'AI-mania' critique to hit the front page during a real market-cool signal.
  • Direct thematic pair with Neil Rimer's 'money coming back out' and Zuckerberg's 'agents haven't progressed' framings.
  • Sets the community-tone baseline for the H2 2026 AI-critique conversation.
communitypolicy
Source →

OpenAI reduces Codex context 372k → 272k; Claude Code runs Bun-in-Rust

OpenAI merged a change reducing the Codex model context from 372k to 272k tokens — a rare tier walkback. Simon Willison documents that Claude Code now runs Bun written in Rust — a runtime consolidation practitioners will note.

News Hacker News

OpenAI reduces Codex Model Context Size from 372k to 272k

OpenAI Codex PR: context reduced from 372k to 272k tokens.

context (was) 372,000context (now) 272,000
Why it matters
  • Rare tier walkback on a widely-used product feature — direct signal about backend cost or capability constraints.
  • Compounds the GPT-5.5 Codex reasoning-token-clustering regression and GPT-5.6 Sol file-deletion reports.
  • Materially affects the Ploy 2.2x-faster / 27%-cheaper migration story and enterprise procurement models.
  • Sets a procurement red flag that model-tier headline capabilities can be silently narrowed.
productscodemarket
Source →
News Hacker News

Claude Code uses Bun written in Rust now

Simon Willison (85 HN points): Claude Code has moved to a Bun runtime written in Rust.

Why it matters
  • Runtime consolidation for the coding-agent surface — practical enterprise-procurement question about supply chain.
  • Compounds the Claude Code 33k-token overhead teardown and steganography reveal.
  • Sets a reference for the Bun-in-Rust pattern other agents may adopt.
codeproductstools
Source →

Dave Eggers at OpenAI: 'silencing a generation'; Kimi threat or menace

The Verge: Dave Eggers told OpenAI staff at their own event that ChatGPT is 'silencing an entire generation.' TC published 'Kimi: Threat or menace?' as the Moonshot Kimi 3 story from Thursday keeps compounding.

News The Verge AI

Dave Eggers told OpenAI staff that ChatGPT was 'silencing an entire generation'

The Verge: Dave Eggers publicly challenged OpenAI staff — at their own event — that ChatGPT is silencing a generation.

↳ Linked News AI Mania Is Eviscerating Global Decision-Making
Why it matters
  • Rare on-site cultural critique of OpenAI by a marquee literary voice.
  • Direct thematic pair with the AI-mania essay and Wired's 'stop making me opt out' piece.
  • Sets a cultural counterpoint to OpenAI's family-focused rollout and 'teens deserve safe AI' framework.
communityproductspolicy
Source →
News TechCrunch AI

Kimi: Threat or menace?

TC essay assessing Kimi 3's threat level to the US frontier as the largest Chinese open model draws near.

Why it matters
  • Framing piece for the largest incoming Chinese open-weights model — clarifies the competitive stakes.
  • Direct follow-through to the Kimi-3-vs-Opus-4.8 FT report and Delangue's 'real race isn't at the frontier' thesis.
marketopen-weightscommunity
Source →

Papers

3 items

DeepLoop, UniVR, Token Time Continuous Diffusion

Weekend HF papers: DeepLoop scales depth in looped transformers; UniVR unifies visual reasoning; Token Time Continuous Diffusion pushes diffusion-based language modeling.

Also today