AI Watchtower Briefing — 2026-07-17

🔴 High Significance

Developer Tools

🔴 💬 Prism accidentally leaked [D] — score 94 Sources: reddit/r/MachineLearning

Just found out from Prism's Discord that compiling is returning someone else's paper. There's a Twitter post too. https://x.com/JustanOthRando/status/2078169169267482778?s=20 I commend their prompt response, though. They took the websit

🔴 💬 How do you actually keep up with AI dev tools/techniques without drowning in noise? — score 94 Sources: reddit/r/AIAgents

Hey there, Like a lot of people, I think AI is a huge opportunity right now, so I'm trying to keep up. I follow people on X, LinkedIn, some subreddits, YouTube channels, etc. But I'm not happy with the signal-to-noise ratio. Most of what I see is clickbait with zero value: "Anthropic vs OpenAI vs X,

🔴 💬 Chinese President Xi Jinping speaks at World AI Conference and reaffirms commitment to open source to promote"openness and win-win" — score 88 Sources: reddit/r/LocalLLaMA

From Vincent Chow on 𝕏: https://x.com/vince_chow1/status/2077947375964791028

🔴 💬 Has anyone actually completed a purchase with their own agent, end to end? Where does it break for you? — score 72 Sources: reddit/r/AIAgents

I hate shopping, so I've been building an agent to do it for me, and I've hit a wall that I can't tell is my problem or everyone's. Discovery and cart-building work fine. The agent finds the thing, compares options, gets to checkout. It's the last few inches that fall apart. Either it's grinding thr

🔴 🐙 rohitg00/ai-engineering-from-scratch — Learn it. Build it. Ship it for others. — score 70 Sources: github_trending

Learn it. Build it. Ship it for others.

Infrastructure & Compute

🔴 💬 Anyone else completely tuning out these massive "open weight" drops? — score 81 Sources: reddit/r/LocalLLaMA

Tbh the benchmarks on stuff like GLM-5.2 look insane. 753B params, 1M context, MIT license... everyone is throwing a party on the front page right now. But like... what is actually "local" about this anymore? A 700B+ MoE isn't fitting on anyone's home rig. Even if you absolutely crush it down to a q

Research Papers

🔴 🤗 LongStraw: Long-Context RL Beyond 2M Tokens under a Fixed GPU Budget — score 82 Sources: huggingface · arxiv/cs.LG

A growing gap separates inference context lengths from RL post-training: inference systems are approaching million-token contexts, while post-training workloads often remain at 256K tokens or below and rely on length generalization at deployment. The gap is especially important for AI agents, whose

🔴 🤗 VIABench: A Comprehensive Video Benchmark Collected from Blind Individuals for Visual Impairment Assistance — score 75 Sources: huggingface

Visually impaired individuals (VIIs) encounter significant daily challenges due to limited access to visual information. Although Multimodal Large Language Models (MLLMs) have achieved impressive results on general vision and language tasks, their practical utility in real-world blind assistance sti

Other Signals

🔴 🧡 Kaiser nurses say AI, workplace surveillance are making their jobs, care worse — score 83 Sources: hackernews

🔴 💬 Kimi K3 is top of nextjs eval — score 73 Sources: reddit/r/LocalLLaMA

🟡 Notable

Model Releases

🟡 💬 Stereo2Spatial: Convert Stereo Music Tracks to Spatialized Binaural Mixes [P] — score 69 Sources: reddit/r/MachineLearning

I have released a model that I have been working on for ~6 months off and on. I've been enjoying listening to spatial music, but there is a lot of music out there with no real quality spatial mix. So I decided to make a model to convert stereo to spatial. I started by making a flow-matching diffusi

🟡 💬 Trellis.cpp now produces high quality assets — score 65 Sources: reddit/r/LocalLLaMA

Some of you might remember that I posted some time ago about the GGML-ported asset production pipeline. A key elelent of that was the TRELLIS.2 port that performs image-to-3D generation. Well, I'm

🟡 ✉️ THE SAME TYPESCRIPT COSTS 73% MORE TOKENS ON CLAUDE THAN GPT (7 MINUTE READ) — score 65 Sources: newsletter/tldr

🟡 ✉️ GPT-5.6 IS NOW AVAILABLE IN FIGMA MAKE (3 MINUTE READ) — score 65 Sources: newsletter/tldr

🟡 ✉️ APPLE DEVELOPING NEW APPLE PENCIL MODELS FOR RELEASE NEXT YEAR, POSSIBLY WITH REPLACEABLE BATTERIES (1 MINUTE READ) — score 65 Sources: newsletter/tldr

Omitted 9 additional model releases items from the main section; see raw data and source-specific sections below.

Developer Tools

🟡 ✉️ FIVE YEARS BUILDING THE CONTEXT LAYER - AND WHY IT BELONGS IN THE AGENTIC LOOP (24 MINUTE READ) — score 65 Sources: newsletter/tldr

🟡 ✉️ WITH AI AGENTS, ATTENTION IS THE NEW BOTTLENECK (2 MINUTE READ) — score 65 Sources: newsletter/tldr

🟡 ✉️ HOW I CUT AN AI AGENT'S TOKEN USE BY 94% (7 MINUTE READ) — score 65 Sources: newsletter/tldr

🟡 ✉️ AI VIDEO INFRASTRUCTURE (WEBSITE) — score 65 Sources: newsletter/tldr

🟡 ✉️ AI VIDEO GENERATOR AND IMAGE GENERATOR (WEBSITE) — score 65 Sources: newsletter/tldr

Omitted 12 additional developer tools items from the main section; see raw data and source-specific sections below.

Infrastructure & Compute

🟡 ✉️ Y Combinator partner and entrepreneur Tom Blomfield announced that he is joining Anthropic as part of the company’s compute team. — score 65 Sources: newsletter/rundown-ai

🟡 ✉️ Meta said its Hyperion data center in Louisiana is expanding to 5 gigawatts of compute, with its new $50B expected investment doubling October's estimate. — score 65 Sources: newsletter/rundown-ai

🟡 ✉️ New York stalls the AI data center boom — score 65 Sources: newsletter/rundown-ai

🟡 ✉️ Why it matters: Data centers have become a massively polarizing AI issue in the U.S., with growing anger (some founded, some inflamed) across every buildout. It's part of why Elon Musk and others are pushing for data cen — score 65 Sources: newsletter/rundown-ai

🟡 🏢 A scorecard for the AI age — score 50 Sources: lab_blog/OpenAI

Sarah Friar, CFO of OpenAI, introduces a practical AI scorecard to measure ROI through useful work, cost per successful task, dependability, and return on compute.

Business & Funding

🟡 💬 ACL ARR 2026 - can't seem to find the review issue report button anywhere? [D] — score 44 Sources: reddit/r/MachineLearning

hey, first time going through this process, the deadline is by today AoE so i'm kinda worried, i just can't seem to find it next to the "official comment" button on my submission, what am i doing wrong? anyone who's having the same issue? https://preview.redd.it/lywderhn4rdh1.png?width=1980&form

Research Papers

🟡 🤗 AsySplat: Efficient Asymmetric 3D Gaussian Splatting for Long-Sequence Scene Modeling — score 60 Sources: huggingface

Recent generalizable 3D Gaussian Splatting models have advanced long-sequence novel view synthesis (NVS), but at the cost of substantial redundant computation. We identify that the redundancy can be mitigated based on two observations: (i) high-precision geometry is not strictly required for high-qu

🟡 🤗 Token Time Continuous Diffusion for Language Modeling — score 42 Sources: huggingface · arxiv/cs.CL

In this paper we introduce token time continuous diffusion (TTCD), a new diffusion language model which (a) operates in continuous space, deterministically mapping Gaussian noise to a final token canvas with no further sampling, and crucially (b) incorporates a new notion of per-token times, with so

🟡 🤗 RxBrain: Embodied Cognition Foundation Model with Joint Language-Visual Reasoning and Imagination — score 40 Sources: huggingface

Embodied cognition requires agents to connect high-level task reasoning with the physical states to be achieved. We introduce Hy-Embodied-RxBrain, an embodied cognition foundation model with joint language-visual reasoning and imagination. Unlike vision-language models that emphasize scene understan

🟡 🤗 SUFLECA: Scaling Up Feature Learning for CAD-to-image Alignment — score 40 Sources: huggingface

CAD-to-image alignment aims to estimate an object's 9D pose (rotation, translation, and anisotropic scale) from a single RGB image, enabling applications in robotics and augmented reality. Recent zero-shot methods use visual foundation models to match image regions to CAD models, yet typically their

Other Signals

🟡 ✉️ HOW TO COLLECT PRODUCT FEEDBACK WHEN YOUR AI GIVES EVERY USER A DIFFERENT ANSWER (3 MINUTE READ) — score 65 Sources: newsletter/tldr

🟡 ✉️ IOS 27 PUBLIC BETA IS HERE WITH SIRI AI, IPHONE SPEED UPGRADES, AND MORE (9 MINUTE READ) — score 65 Sources: newsletter/tldr

🟡 ✉️ APPLE'S LAWSUIT THREATENS TO DISRUPT OPENAI'S BID TO RIVAL THE IPHONE (10 MINUTE READ) — score 65 Sources: newsletter/tldr

🟡 ✉️ CLOUDFLARE GIVES OPENAI NETWORK SIGNALS COVERING 20% OF THE WEB (10 MINUTE READ) — score 65 Sources: newsletter/tldr

🟡 ✉️ AVOID THE AI EXPERTISE TRAP (2 MINUTE READ) — score 65 Sources: newsletter/tldr

Omitted 13 additional other signals items from the main section; see raw data and source-specific sections below.

🟢 Incremental

Model Releases

🟢 💬 Gemma4-31b better than Qwen3.6-27b — score 35 Sources: reddit/r/LocalLLaMA

So I know this is going to incur the wrath of the Qwen cult but after a month of using 27b q8_0 as my primary coding agent in 6+ agent coding workflow with GPT-5.5 as the orchestrator, I got very frustrated with the amount of back and forth I was doing just to always end up in the same place withou

🟢 💬 [RESEARCH] Breaking the 1-bit Floor: Achieving "Negative-Bit Quantization" (NBQ) via Phase-Inverted Tensor Embedding (satire) — score 27 Sources: reddit/r/LocalLLaMA

Hey everyone, I’ve spent the last three weeks compiling custom llama.cpp forks and running imatrix maps on a modified CUDA kernel setup, and the numbers don’t lie. We’ve been looking at model compression completely wrong. Everyone in the community has assumed that 1-bit quantization (like BitNet o

🟢 💬 User experience of Bonsai-Ternary-27B on 4060Ti 16GB for KB management and productivity assistant use cases — score 19 Sources: reddit/r/LocalLLaMA

Howdy folks, Long time reader here. This time, I would like to share my experience with this model that I have been cautiously optimistic about when I saw the announcement post on this thread. ----- My setup: * AMD Ryzen 5 on AM5 chipset (I genuinely don't remember the exact CPU. Just pick what

🟢 🧡 Homomorphically encrypted CIFAR-10 inference in 200ms — score 17 Sources: hackernews

🟢 💬 When will we get more small LLMs? — score 12 Sources: reddit/r/LocalLLaMA

Basically the title. We had our last drop in the beginning of April, do we just not get a refresh of Gemma or Qwen?

Developer Tools

🟢 💬 What boring piece of infrastructure became unexpectedly important once you started putting agents into production? — score 11 Sources: reddit/r/AIAgents

Most conversations about building agents seem to revolve around models, prompts, RAG, memory and tool calling. Then you try to get one running reliably outside of a local setup and suddenly you're dealing with retries, queues, permissions, logging, rate limits, tracing, rollbacks and a bunch of othe

🟢 🐙 vercel-labs/just-bash — Bash for Agents — score 11 Sources: github_trending

Bash for Agents

🟢 🐙 sourcebot-dev/sourcebot — Sourcebot is a self-hosted tool that helps humans and agents understand your codebase. — score 7 Sources: github_trending

Sourcebot is a self-hosted tool that helps humans and agents understand your codebase.

Research Papers

🟢 🤗 Chat2Scenic: An Iterative RAG-Based Framework for Scenario Generation in Autonomous Driving — score 15 Sources: huggingface

Validating autonomous driving systems requires diverse, regulation-compliant test scenarios. In simulation-based testing, scenarios are defined as executable scripts. Yet automatically generating such scripts from regulatory descriptions remains an open challenge, and existing approaches face fundam

🟢 🤗 Hierarchical Denoising For Multi-Step Visual Reasoning — score 15 Sources: huggingface

Video models are evolving into vision foundation models, yet they still lack human-like multi-step reasoning. Streaming autoregressive diffusion models are efficient but limited in reasoning, while bidirectional diffusion enables global revision with high inference costs due to dense frame-level den

Other Signals

🟢 💬 EU AI Act OpenRAG: 933 legally structured chunks and BGE-M3 embeddings in one SQLite file [P] — score 26 Sources: reddit/r/MachineLearning

I have released EU AI Act OpenRAG, a downloadable corpus of Regulation (EU) 2024/1689 designed for RAG and legal-NLP experimentation. Instead of sliding character windows, the corpus chunks on the Regulation’s legal structure: * one chunk per article paragraph * one per recital * one per Article 3 d

🟢 💬 short-paper at ACL/EMNLP/EACL [R] — score 25 Sources: reddit/r/MachineLearning

Does anyone have accepted short-paper at ACL/EMNLP/EACL 2025/26? Could you share your track and overall assessment? I'm just trying to get a sense of things, as it seems short papers have a lower acceptance rate than long ones.

🟢 💬 TACL journal doubts [D] — score 25 Sources: reddit/r/MachineLearning

I submitted my TACL paper approx on June 1th and was scheduled for July 1st cycle, when and how do you guys think we'll be getting our reviews given the July cycle for the paper which I've submitted at TACL ? And how long does the entire process take for those who have submitted to TACL ? Also, I do

🟢 💬 I built an n8n workflow that turns any topic into a research report automatically — score 24 Sources: reddit/r/AIAgents

I've been building a few automation workflows with n8n lately, and this is one I'm pretty happy with. The workflow starts with a single topic. From there it automatically: * Generates multiple research queries with OpenAI * Pulls information from Google Search, Wikipedia, NewsAPI, and Google Scholar

🟢 💬 BMVC rebuttals update [D] — score 6 Sources: reddit/r/MachineLearning

Rebuttal access opened to reviewers on July 11 (19:05 UTC), so any later modification means final score updated (even if it's hidden from us now). It shows like this over each review: Official Review by Reviewer KBVi 22 Jun 2026, 18:17 IST (modified: 17 Jul 2026, 17:39 IST) How many of your reviews

Omitted 1 additional other signals items from the main section; see raw data and source-specific sections below.

📈 Trending Repos

Repo	Description	Stars Today	Language
rohitg00/ai-engineering-from-scratch	Learn it. Build it. Ship it for others.	218	python
anthropics/cwc-workshops		37	typescript
vercel-labs/just-bash	Bash for Agents	6	typescript
sourcebot-dev/sourcebot	Sourcebot is a self-hosted tool that helps humans and agents understand your codebase.	5	typescript

📄 New Papers

Title	Category	Hotness	Link
LongStraw: Long-Context RL Beyond 2M Tokens under a Fixed GPU Budget	research_paper	42	Open
VIABench: A Comprehensive Video Benchmark Collected from Blind Individuals for Visual Impairment Assistance	research_paper	8	Open
AsySplat: Efficient Asymmetric 3D Gaussian Splatting for Long-Sequence Scene Modeling	research_paper	5	Open
Just Keep Prompting: Evaluating Repetitive Socratic Prompting in VLMs	cs.CL	0	Open
Quantum Compositional NLP for Arabic: Grammar, Morphology, and Word Sense in Circuit Topology	cs.CL	0	Open
LBA: Textual Hard-Label Adversarial Attack under Low Query Budgets	cs.CL	0	Open
UniSAGE: Unifying Static and Dynamic Attributes with Hyper-Structure	cs.CL	0	Open
Latent Communication Between Language Model Agents: Channels, Alignment, and the Limits of Text	cs.CL	0	Open
UzWordnet and Generative AI for Learning Uzbek by Game Playing	cs.CL	0	Open
Automatically Evolving Prompt Guidelines for Task-Specific Optimization	cs.CL	0	Open
Polestar: Drift-Aware Cache Calibration and Token Commitment for Efficient Inference of Diffusion LLMs	cs.CL	0	Open
Eta Given Delta: Defining LLM Tool Efficiency With Marginal Tool Utility	cs.CL	0	Open
Simplicity Paradox: Debunking myths about prompting and datasets for LLM evaluation	cs.CL	0	Open
MAPS: Modeling Co-Existing Subjective Perspectives and Shared Meaning in Multi-Agent Cognitive Dialogue	cs.CL	0	Open
Introspection Fine-Tuning (IFT): Training Small LLMs to Introspect	cs.CL	0	Open

🏢 Lab Blog Posts

• OpenAI: A scorecard for the AI age

🐦 Twitter/X Highlights

Account	Tweet Summary
OpenAI	GPT-5.6 Sol sets a new state of the art in cybersecurity on “The Last Ones” cyber range. We’re already seeing that capability translate into defensive outcomes: helping teams find, validate, and fix vulnerabilities in real-world code. Put it to work with Codex Security: https://openai.com/daybreak/c Post
OpenAI	Pinned: 10,000 reasons people love GPT-5.6 Sol https://switch-to-codex.openai.chatgpt.site/ Post

Newsletter

• tldr: FIVE YEARS BUILDING THE CONTEXT LAYER - AND WHY IT BELONGS IN THE AGENTIC LOOP (24 MINUTE READ)
• tldr: HOW TO COLLECT PRODUCT FEEDBACK WHEN YOUR AI GIVES EVERY USER A DIFFERENT ANSWER (3 MINUTE READ)
• tldr: WITH AI AGENTS, ATTENTION IS THE NEW BOTTLENECK (2 MINUTE READ)
• tldr: IOS 27 PUBLIC BETA IS HERE WITH SIRI AI, IPHONE SPEED UPGRADES, AND MORE (9 MINUTE READ)
• tldr: APPLE'S LAWSUIT THREATENS TO DISRUPT OPENAI'S BID TO RIVAL THE IPHONE (10 MINUTE READ)
• tldr: CLOUDFLARE GIVES OPENAI NETWORK SIGNALS COVERING 20% OF THE WEB (10 MINUTE READ)
• tldr: HOW I CUT AN AI AGENT'S TOKEN USE BY 94% (7 MINUTE READ)
• tldr: THE SAME TYPESCRIPT COSTS 73% MORE TOKENS ON CLAUDE THAN GPT (7 MINUTE READ)
• tldr: AVOID THE AI EXPERTISE TRAP (2 MINUTE READ)
• tldr: GPT-5.6 IS NOW AVAILABLE IN FIGMA MAKE (3 MINUTE READ)
• tldr: APPLE DEVELOPING NEW APPLE PENCIL MODELS FOR RELEASE NEXT YEAR, POSSIBLY WITH REPLACEABLE BATTERIES (1 MINUTE READ)
• tldr: WE DIDN'T TEACH OUR DESIGNERS AI: WE BUILT A PLACE WHERE THEY COULD LEARN IT TOGETHER (7 MINUTE READ)
• tldr: AI VIDEO INFRASTRUCTURE (WEBSITE)
• tldr: AI VIDEO GENERATOR AND IMAGE GENERATOR (WEBSITE)
• tldr: THE CONTEXT LAYER FOR AI: BRINGING TRUSTED DROPBOX CONTENT INTO OPENAI WORKFLOWS (4 MINUTE READ)
• tldr: APPLE SAYS FORMER EMPLOYEE EXPLOITED ‘RARE' BUG TO DOWNLOAD CONFIDENTIAL FILES AFTER LEAVING FOR OPENAI (4 MINUTE READ)
• tldr: GOOGLE CLOUD EXPANDS PARTNERSHIP WITH SAMSUNG TO ACCELERATE AGENTIC AI (2 MINUTE READ)
• tldr: GHOSTCOMMIT' HIDES PROMPT INJECTION IN IMAGES TO FOOL AI AGENTS AND STEAL SECRETS (2 MINUTE READ)
• tldr: ACCELERATING EDR EVASION WITH LLM-DRIVEN ANALYSIS (17 MINUTE READ)
• tldr: HOW MICROSOFT SHIPS THOUSANDS OF PRODUCTION AI AGENTS (18 MINUTE READ)
• tldr: VIDEO GENERATORS AS GENERAL-PURPOSE VISION MODELS (8 MINUTE READ)
• tldr: TESTING AGENTS ON LONG-HORIZON TERMINAL WORK (GITHUB REPO)
• tldr: OPEN-WEIGHT MODELS REACHED 29% OF AI GATEWAY USAGE (9 MINUTE READ)
• tldr: TOM BLOMFIELD IS TAKING A LEAVE OF ABSENCE FROM YC TO JOIN ANTHROPIC (1 MINUTE READ)
• tldr: THE AGENTIC ECONOMY TREATISE (WEBSITE)
• tldr: OPENAI'S FIRST DEVICE WILL BE MOVABLE, SCREENLESS SPEAKER BUILT AS AI COMPANION (9 MINUTE READ)
• rundown-ai: Anthropic expanded Fable 5’s availability in Claude paid plans for a third time, with the model now accessible through July 19 before switching to usage credits.
• rundown-ai: Microsoft CEO Satya Nadella published an essay titled "The Reverse Information Paradox," arguing firms pay for AI twice, in money and in proprietary knowledge.
• rundown-ai: Y Combinator partner and entrepreneur Tom Blomfield announced that he is joining Anthropic as part of the company’s compute team.
• rundown-ai: Turing Award winner Richard Sutton co-founded Oak Lab, a new AI company building “animal-like intelligence that learns purely from its own experience”.
• rundown-ai: Meta said its Hyperion data center in Louisiana is expanding to 5 gigawatts of compute, with its new $50B expected investment doubling October's estimate.
• rundown-ai: DeepMind CEO pitches U.S. AI watchdog
• rundown-ai: Report: OpenAI's first device is an AI speaker
• rundown-ai: Capabilities include fielding questions, sending replies, queuing music, and operating smart-home gear via GPT-Live, OAI’s recently upgraded voice mode.
• rundown-ai: Build a no-code voice agent for customer intake
• rundown-ai: 1. Go to Grok’s Voice Agent Builder and hit Create a "Lead Qualification" agent. Then, describe your business, its target, and what a good lead should answer
• rundown-ai: New York stalls the AI data center boom
• rundown-ai: Why it matters: Data centers have become a massively polarizing AI issue in the U.S., with growing anger (some founded, some inflamed) across every buildout. It's part of why Elon Musk and others are pushing for data cen
• rundown-ai: Mercury 2 - Inception's diffusion reasoning model for realtime voice agents
• rundown-ai: Claude for Teachers - Free Claude with teaching and lesson-planning skills
• Latent Space: The cloud was built for developers. Butagents are now changing that. - first seen 2026-07-16
• Latent Space: At the time, Modal was just a teeny little company with a$17M Series A. - first seen 2026-07-16
• Latent Space: Today, fresh off their$355M Series C, Modal is one of the clearest examples of the agent cloud future being built in real time: a cloud platform moving past traditional web app assumptions toward the - first seen 2026-07-16
• tldr: APPLE IN TALKS WITH STARTUP THAT SHRINKS AI MODELS TO RUN ON AN IPHONE (9 MINUTE READ) - first seen 2026-07-15

Repeated From Recent Briefings

• Graphify-Labs/graphify — AI coding assistant skill (Claude Code, Codex, OpenCode, Cursor, Gemini CLI, and more). Turn any folder of code, SQL schemas, R scripts, shell scripts, docs, papers, images, or videos into a queryable knowledge graph. App code + database schema + infrastructure in one graph. - first seen 2026-06-26
• KIMI K3 Beats Claude Fable and GPT 5.6 sol in arena.ai!!! - first seen 2026-07-16
• Shubhamsaboo/awesome-llm-apps — 100+ AI Agent & RAG apps you can actually run — clone, customize, ship. - first seen 2026-07-11
• lobehub/lobehub — 🤯 LobeHub is your Chief Agent Operator, organizing your agents into 7×24 operations by hiring, scheduling, and reporting on your entire AI team. - first seen 2026-07-16
• PostHog/posthog — 🦔 PostHog is the leading platform for building self-driving products. Our developer tools – AI observability, analytics, session replay, flags, experiments, error tracking, logs, and more – capture all the context agents need to diagnose problems, uncover opportunities, and ship fixes. Steer it all from Slack, web, desktop, or the MCP. - first seen 2026-07-14
• Spectral Rewiring for Exploration, Purification, and Model Merging - first seen 2026-07-07
• openinterpreter/openinterpreter — A coding agent for open models like Kimi K3 - first seen 2026-06-15
• whats the best and complete way to keep up with ai/ml news? [D] - first seen 2026-07-16
• anomalyco/opencode — The open source coding agent. - first seen 2026-05-09
• openai/codex — Lightweight coding agent that runs in your terminal - first seen 2026-06-24
• ... plus 319 more repeated items in processed data