Pier 99

Steph

A short daily briefing on the most interesting stories in AI, hosted by Steph. pier99.substack.com

  1. Aug 25

    Episode 13: Codex Brings Back the Five-Hour Limit

    In this episode of Pier 99: * OpenAI brings back the five-hour usage window for Plus users across ChatGPT Work and Codex, while Tibo promises the biggest DevDay in company history. * Grok Voice Think Fast 2.0 reaches No. 1 on Artificial Analysis, as Starlink says it already handles more than 15,000 support and sales calls a day. * ByteDance launches Doubao Work, an AI office Agent that can create documents, research topics, manage files, and operate websites. * Wan 3.0 lands on Runway, Pika, and Topview, with video platforms competing on price and unlimited-generation plans. * ChatGPT Images adds transparent backgrounds and custom sticker packs for iMessage and WhatsApp. Links and sources OpenAI and CodexCodex five-hour limit · Community reaction · OpenAI DevDay 2026 tease Grok Voice and StarlinkSpeech-to-speech ranking · Starlink call and order volume · Official Grok Voice announcement Doubao WorkOfficial product page · Launch report · ByteDance product consolidation discussion Wan 3.0Runway availability · Pika API Club examples · Topview price comparison ChatGPT ImagesStickers and transparent backgrounds This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit pier99.substack.com

    Episode 13: Codex Brings Back the Five-Hour Limit

  2. Aug 15

    Episode 12: Cloudflare Ships Fast. Can Yila Outrun Codex?

    Cursor has officially joined SpaceX. Its team will join SpaceXAI with a practical mission: help make Grok the world’s most useful AI. Cloudflare also deserves some attention. In just two months, the company published more than 60 AI-related posts and shipped more than 20 AI and agent products or features. It is building the computers, browsers, security controls, and payment rails that agents will need. Then there is Yila.ai, “the Codex for research.” It can find real papers, write sourced literature reviews, analyze data, and generate academic posters and slides. Explore Yila.ai — The Codex for research Yila looks useful, but it also exposes the question hanging over every vertical AI startup: what happens if Codex or Claude turns your workflow into a plugin or a few Skills? Paul Graham’s recent comments suggest one possible answer. As software moats get thinner, founders may have to build deeper technology and enter harder industries. Links and sources * Cursor officially joins SpaceX * SpaceXAI welcomes Cursor * Artificial Analysis evaluates Grok 4.6 * Cloudflare introduces an Agent computer runtime * Cloudflare introduces WebMCP * Cloudflare introduces Kitesurf * Cloudflare unifies Workers AI and AI Gateway * Cloudflare adds MCP traffic detection * Cloudflare Agents SDK x402 payments * Cloudflare Monetization Gateway * Yila.ai — The Codex for research * Yila private-beta examples * Paul Graham on today’s YC startups * Paul Graham says the change is mostly the market This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit pier99.substack.com

    Episode 12: Cloudflare Ships Fast. Can Yila Outrun Codex?

  3. Aug 14

    Episode 11: GLM-5.3 and Gemini 3.7 Flash Launch. DeepSeek Prices Jump.

    AI model releases are speeding up again. In this episode, Steph covers GLM-5.3’s coding and cybersecurity gains, including Z.ai’s claim that the model found 2,436 real vulnerabilities across 269 open-source projects. We also look at Gemini 3.7 Flash’s improvements in coding, web development, reasoning, speed, and temporary discounted pricing; OpenAI’s GPT-5.6 Sol Ultrafast preview; DeepSeek V4 Pro’s new peak pricing; the open-source DeepSeek Harness for building AI agents; and ChatGPT’s new optional Computer History feature. Links and sources GLM-5.3 * Z.ai announcement * Coding benchmark details * Cybersecurity and open-weights note Gemini 3.7 Flash * Google DeepMind announcement * Google product overview * Pricing and availability * Artificial Analysis evaluation GPT-5.6 Sol Ultrafast * OpenAI preview DeepSeek V4 Pro pricing and DeepSeek Harness * DeepSeek V4 Pro announcement * Official API pricing * DeepSeek Harness announcement * DeepSeek Harness on GitHub ChatGPT Computer History * OpenAI announcement * Timeline and privacy controls * Skills and Scheduled Tasks details This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit pier99.substack.com

    Episode 11: GLM-5.3 and Gemini 3.7 Flash Launch. DeepSeek Prices Jump.

  4. Aug 13

    Episode 10: Grok in Heaven, DeepSeek in Hell

    Two major models landed almost at once, but the reactions could not have been more different. Grok 4.6 reached third place on the Artificial Analysis Intelligence Index and undercut GPT-5.6 Sol on price, while DeepSeek V4 Pro arrived through a confusing rollout and underwhelming early reactions. Just before publication, DeepSeek finally made the release official, so we’ll revisit it after testing. Also in this episode: Codex crossed 15 million active users and got another reset, Gemini added 14 new app connections, and Claude in Chrome became part of Claude’s cross-device Agent system. Links and sources DeepSeek V4 Pro Official announcement: Arena release: Arena details: Artificial Analysis comparison: https://artificialanalysis.ai/models/comparisons/deepseek-v4-flash-vs-deepseek-v4-pro Grok 4.6 and Grok 4.7 SpaceXAI announcement: SpaceXAI details: SpaceXAI availability: Artificial Analysis ranking: Arena ranking: Elon Musk’s Grok 4.7 comments: Codex reaches 15 million active users Gemini adds 14 app connections Official announcement: OpenTable: Ticketmaster and Fever: GetYourGuide and Localiza: Zocdoc, Thumbtack, and Angi: Granola, Otter.ai, and Zoho: Wix, iHeartRadio, and Pandora: Rollout details: Claude in Chrome goes cross-device Official announcement: Conversation history: Browser safety guidance: This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit pier99.substack.com

  5. Aug 12

    Episode 9: ChatGPT Gets New Features, Gemini Hits 1 Billion Monthly Users

    Today on Pier 99: Tibo revives his Codex reset promise and teases a mystery surprise. OpenAI also brings ChatGPT and Codex to Linux, while making it easier to import projects, chats, Skills, and Plugins from other agents. Tencent puts AI at the center of its second-quarter story, with Hy3, the planned Hy4, office agents, and Weixiao inside WeChat. Gemini reaches one billion monthly users as voice, camera, screen sharing, and image generation reshape how people use it. Anthropic prepares invisible text watermarks and a Claude detection API. And Grok Bot gives AI teammates their own computers, raising a simple question: how much access will users actually trust an agent with? Links and sources Codex surprise, Linux, and agent importsTibo teases a Codex surpriseOpenAI announces ChatGPT for LinuxSupported Linux distributionsOpenAI announces agent imports and sync Tencent makes AI a core focusTencent AI’s Q2 updateTencent Investor Relations Gemini reaches one billion usersSundar Pichai announces the milestoneGoogle’s Gemini usage report Claude text watermarkingThariq explains the watermark and detection APICoverage and limitations Grok BotGrok Bot early-beta announcementLenny Rachitsky’s early experience This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit pier99.substack.com

    Episode 9: ChatGPT Gets New Features, Gemini Hits 1 Billion Monthly Users

  6. Aug 11

    Episode 8: OpenAI Moves Into Cybersecurity

    Tibo is everywhere. Jason’s first-class joke pulled Thariq into another OpenAI–Anthropic exchange, while ZCode celebrated one million users by resetting GLM Coding Plan limits. OpenAI introduced GPT-5.6-Cyber and opened applications for Daybreak Red, pushing its models deeper into professional vulnerability research and red teaming. Anthropic says an unreleased Claude made a dramatic leap on a result related to the Riemann hypothesis. Claude Sonnet 5’s introductory API price is now permanent, and MAI-Image-2.6 quietly reached No. 2 on Arena before Microsoft gave it a proper launch. Links and sources Tibo, Jason, and Thariq Jason’s original post · Thariq’s Tibo reply · Jason’s business-class reply ZCode reaches one million users ZCode usage reset GPT-5.6-Cyber and Daybreak OpenAI announcement · Vulnerability research and Chrome V8 · Access restrictions and safeguards Claude and the Riemann hypothesis Anthropic announcement · Jarred Sumner’s experiment · Levinson’s 1974 result · The five-twelfths result Claude Sonnet 5 pricing Permanent Sonnet 5 pricing · Unlimited GPT-5.6 Luna text chat MAI-Image-2.6 Arena ranking · Category results · Mustafa Suleyman’s reaction This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit pier99.substack.com

  7. Aug 10

    Episode 7: Meta's 30B Agent Specialist

    Sam Altman joins the Tibo-Boris fun. Meta launches Muse Glimmer, a 30-billion-parameter open-weight model specialized for Agent workflows. Alibaba turns Qwen into a service platform that can complete real-world transactions. And Unitree opens its IPO to investors, with demand expected to be intense. Links and sources Sam Altman praises TiboSam Altman on X Meta Muse GlimmerIntroductionLocal optimizationAgent workflow demoDownloads and technical resources Qwen Open PlatformFirst 12 service AgentsCross-device platform summaryAlibaba’s official Qwen product overviewAlibaba’s official Agentic AI strategy Unitree IPOOffer price and subscription updateCSRC registration approvalOfficial SSE prospectusSubscription schedule and codes This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit pier99.substack.com

    Episode 7: Meta's 30B Agent Specialist

  8. Aug 9

    Episode 6: Anthropic Tried to Hire OpenAI’s Codex Lead—In Public

    A developer followed Tibo’s tutorial for running an OpenAI model inside Claude Code—and shortly afterward, Anthropic suspended his account for “suspicious signals.” Claude Code lead Boris Cherny stepped in, explained that an account classifier had probably misfired, and then publicly offered Tibo a job at Anthropic. Tibo declined, defended “freedom of harness,” and fulfilled an old promise by resetting paid users’ Codex limits. When someone called the reset performative, he promised another performative reset on Monday. Alex’s account was restored, and everyone got a surprisingly entertaining happy ending. Links and sources Tibo’s original Claude Code setup: Alex reports the suspension: Alex’s technical correction and timeline: Tibo responds to the suspension: Tibo promises another “performative reset” on Monday: Boris explains the classifier and publicly recruits Tibo: Tibo responds and turns down the offer: Tibo resets paid-user usage limits: Boris confirms the account should be restored: Alex confirms the happy ending: This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit pier99.substack.com

    Episode 6: Anthropic Tried to Hire OpenAI’s Codex Lead—In Public
See All (13)

About

A short daily briefing on the most interesting stories in AI, hosted by Steph. pier99.substack.com

Information

  • Creator
    Steph
  • Years Active
    2026
  • Episodes
    13
  • Rating
    Clean
  • Copyright
    © Steph
  • Show Website
    Pier 99