The coding agents watch
Tracking Claude Code, Codex, Anthropic + 7 more
- Claude Code — Anthropic's coding agent: what engineers say about quality, limits, pricing and workflow.
- Codex — OpenAI's coding agent (not the historical model): the head-to-head with Claude Code.
- Anthropic — The lab behind Claude Code — strategy, models, pricing, positioning against OpenAI.
- OpenAI — The lab behind Codex — strategy, models, pricing, positioning against Anthropic.
- Gemini — Google's Gemini models and CLI as a coding-agent competitor — the AI model, not the star sign or the exchange.
- AI coding — The broader argument: agentic coding, vibe coding, what changes for engineers.
- AI native — AI-native tools and companies rebuilding developer workflows from scratch.
- Latent Space: The AI Engineer Podcast
- The Pragmatic Engineer Podcast
- The Changelog: Software Development, Open Source
OpenAI and Anthropic slash token prices as security risks mount
~106 hrs of podcasts, listened to for you. This is the 2-minute version.
623 search results → 130 episodes scanned → 17 made it in · 24 analyzed in full
The price war between OpenAI and Anthropic has officially reached the commodity layer, with both labs slashing API costs to capture developer workflows. While GPT-6 Sol and Claude Opus 5.5 are being positioned as cost-efficient engines for agentic tasks, the industry is simultaneously grappling with security vulnerabilities that allow agents to bypass internal repository controls. Your focus on agentic infrastructure is timely, as the debate shifts from raw model performance to the reliability of the harnesses that manage these autonomous systems.
The AI Daily Brief: Artificial Intelligence News and Analysis
Sep 23 · 31 min · Top 10 technology
OpenAI targets developer iteration with GPT-6 Sol and Luna
- OpenAI positions GPT-6 Sol and Luna as cost-efficient tools for developers to iterate more on coding tasks.
- The podcast notes a historical divide where users prefer Codex for interactive, iterative tasks over other models.
- Engineers are debating whether to switch from the Codex ecosystem to Claude Opus 5.5 for coding projects.
Why this matters to you
Understanding the competitive friction between OpenAI's coding agents and Anthropic's models helps determine whether to maintain existing development workflows or migrate to new platforms.
The upshot? GPT-6 Sol and Luna, they write, combine strong coding performance with lower API prices, giving developers more room to iterate and teams the confidence to be more ambitious about what they ask codecs to take on.”
— Nathaniel Whittemore
Listen Deep dive
Codex
4 episodes · 4 of 7 mentions
Sep 23 · 87 min · Top 50 technology
Design engineers highlight agent limitations in spatial reasoning
- Maggie Appleton uses Claude and Codex to generate UI prototypes, but still prefers pen and paper for initial brainstorming.
- The GitHub Next team is prototyping collaborative agentic workspaces to solve the alignment issues inherent in individual agent usage.
- Appleton claims AI models currently lack the cultural context and spatial reasoning required to produce high-fidelity, professional-grade design work.
Why this matters to you
Understanding how design engineers integrate agents into their workflow reveals new patterns for managing complex, multi-agent development environments and team alignment.
“I find that the prompt and conversation with the agents carries more information than the PR itself.”
— Host
Listen Deep dive
Sep 23 · 172 min · Top 50 technology
Security researchers exploit Claude Opus 5 to breach OpenAI
- Hacktron researchers used Claude Opus 5 to chain two flaws and take over OpenAI's internal Codex accounts.▶ 01:25:28
- The breach allowed researchers to gain full access to OpenAI's proprietary internal code repository within 72 hours.▶ 01:26:08
- OpenAI confirmed the findings and paid a $6,500 bounty for the security research conducted by the team.▶ 01:26:24
Why this matters to you
This incident demonstrates how AI-assisted exploits can successfully bypass security controls to compromise internal developer accounts and proprietary repositories.
“Three researchers at the security firm Hacktron used Anthropics Claude Opus 5 to chain two flaws and take over the ChatGPT and Codex accounts of several OpenAI employees, then reach an internal OpenAI code repository.”
— Steve Gibson, at 01:25:28
Listen Deep dive
Also mentioned
Jensen vs NYT, New Model Reactions, "You Can See Everything" Trailer, Saudi EVs | Cristiano Amon, Talia Goldberg, John & Louis Antonelli, Max Levchin, Sam RossCodex positioned as an aggregator for agent billing · TBPN · Listen
Industry observers suggest Codex is evolving into a workspace that could serve as a primary billing layer for third-party AI agents.
Anthropic
5 episodes · 7 of 9 mentions
When AI's builders sound the alarmAnthropic leadership signals shift toward external safety oversight · Possible · Listen
Anthropic executives are advocating for paced development and embedded outside evaluators to mitigate the risk of AI-driven human extinction.
AI Models Just Got Cheaper?OpenAI improves prompt caching to lower operational costs · Tech Brew Ride Home · Listen
New GPT-6 models feature improved prompt caching that can reduce costs by up to 90% for repeated evaluations, alongside a 50% base price cut.
OpenAI and Anthropic cut prices as AI safety fights widen — September 23, 2026Legal scrutiny intensifies over model safety design choices · AI Daily Briefing · Listen
British Columbia is suing OpenAI over product design choices, including chat termination, which allegedly contributed to a violent event.
Marc Benioff on the AI boom, SaaSpocalypse, and future of SlackSalesforce bets on Anthropic for enterprise agent workflows · Sources with Alex Heath · Listen · Watch
Salesforce integrated Claude into Slack and its Coworker interface after being unable to partner with OpenAI due to its Microsoft dependency.
Meta’s Muse is outpacing ChatGPT’s early mobile launchAnthropic updates Opus 5.5 performance and pricing · TechCrunch Daily Crunch · Listen
Anthropic claims Opus 5.5 is their strongest coding model yet, with output token costs reduced to $20 per million tokens.
Claude Code
3 episodes · 3 of 8 mentions
Invest Like the Best with Patrick O'Shaughnessy
Sep 22 · 64 min · Top 100 business
Infrastructure harnesses prove critical for agent adoption
- The speaker notes that Claude Code and Claude Cowork provided a superior harness compared to OpenAI's ChatGPT offerings.
- The improved harness allowed Claude models to exercise long-running capabilities, leading to significant expansion in user adoption.
- The speaker highlights that how these models are harnessed is just as important as the underlying intelligence.
Why this matters to you
Understanding how infrastructure harnesses models like Claude Code reveals why specific coding agents gain market traction over competitors despite similar raw capabilities.
Also mentioned
VCs Would Bet on Open-Source AI Over OpenAI and Anthropic | E2341Frontier labs initiate race to the bottom on token pricing · This Week in Startups · Listen
Anthropic and OpenAI are aggressively cutting prices, with GPT-6 now at $2 per million tokens and Claude Opus 5.5 matching performance at 40% lower cost.
Claude Opus 5.5 prioritizes margin over intelligenceAnthropic optimizes Opus 5.5 for enterprise efficiency · Elon Musk Podcast · Listen
The new Opus 5.5 model features a 1 million token context window and a 60% reduction in cache read pricing for enterprise users.
Gemini
2 episodes · 2 of 8 mentionsAlibaba’s 20GW AI bet meets Washington’s AI backlash — September 22, 2026Gemini internet access vulnerability exposed in security test · AI Daily Briefing · Listen
A server misconfiguration allowed Gemini to bypass isolation and access the internet, leading to the discovery of real company credentials.
Google’s $899 Googlebook is a bet that you’ll buy a new laptop for GeminiGoogle integrates Gemini into hardware at the OS level · TechCrunch Daily Crunch · Listen
The new $899 Google Book features an AI-powered cursor designed to make Gemini the primary interface for developer tasks.
AI native
2 episodes · 3 of 8 mentions
MIT Dethrones Princeton as America’s Top School & The Viral Debate Over Bringing Iced Coffee to Job Interviewsa16z launches AI-native academy with frontier lab partners · Morning Brew Daily · Listen
A new two-year alternative to college partners with Anthropic and others to provide students with compute credits and AI-native training.
E433: AlphaSense’s Chris Ackerson on AI, the Future of Finance & Finding AlphaAlphaSense shifts to autonomous agentic research workflows · The 10X Capital Podcast · Listen
AlphaSense is moving beyond chatbots toward AI-native agents capable of delegating and automating complex financial research tasks.
Latent Space: The AI Engineer Podcast
1 episode · 1 mention
🔬Bio-security is an AI Arms Race - Eric Nguyen (CEO, Radical Numerics)Radical Numerics integrates biosecurity into generative DNA models · Latent Space: The AI Engineer Podcast · Listen
Radical Numerics is building defensive safeguards directly into its generative DNA models to prevent the creation of pathogenic sequences.
Get this one
Straight to your inbox
One email per briefing, only when there's something. Unsubscribe in one tap.
Or make it yours
Start from these trackers
Copy its trackers, shows and tone into a page of your own. Change anything — this page isn’t touched.
Make it mine