#CODEX
今日のZennトレンド

Claude Code / Codexで「私のlimit、減りすぎ…?」と思ったときに見る記事
Claude CodeやCodexの利用上限が早く消費される原因は、ツール実行ごとに全履歴を再送し、キャッシュ分も課金されるためです。
履歴が長くなるほど1リクエストの消費量が増えます。
対策として、セッションを適切に切り替えること、サブエージェントを活用すること、時間を空けて再開する際は新セッションを開始することを推奨しています。
Claude Code / Codexで「私のlimit、減りすぎ…?」と思ったときに見る記事
こんにちは、TOKIUMの花房です。Claude CodeやCodexをサブスクリプションで使っていて、以下のように感じることはないでしょうか。週の半ばで週間上限が尽きる。作業量は先週と変わらないはずなのにSlack botやcron/launchdからエージェントを動かしていて、体感よりもリミット消費が多いプロンプトキャッシュにヒットしているはずなのに、やはり減りが多い気がするCodexの週間上
zenn.dev
September 29, 2026 at 3:20 PM
🚀 Most people use ChatGPT. The ones getting ahead learn to build LLMs.

425 pages, 50 chapters: transformers, tokenizers, fine-tuning (LoRA/QLoRA), RAG, agents, deployment. Zero AI background needed.

👉 www.thebergcodex.shop/products/bui...
Build Your Own LLM (2026) — Train, Fine-Tune & Deploy AI Models | BERG CODEX
Master LLM engineering in 425 pages. Build, train, fine-tune Llama, Mistral & DeepSeek, deploy with Docker & cloud. Beginner to expert. Digital download.
www.thebergcodex.shop
September 29, 2026 at 3:20 PM
Deep in the Shudder Mountains near the village of Steller lies a lost mine. Well, as it turns out, several lost mines. The coal mine has been sealed for decades, but something lurks even deeper...for centuries, perhaps eons. #ttrpg #DCCrpg codex.livetorole.com/products/the...
The Lost Mines of Steller | Codex
Death Guaranteed Games. Deep in the Shudder Mountains near the village of Steller lies a lost mine. Well, as it turns out, several lost mines. The coal mine has been sealed for decades, but something…
codex.livetorole.com
September 29, 2026 at 3:18 PM
OpenAI Dev Day 2026: Live updates on the latest ChatGPT and Codex announcements

#Tech #Gadgets
OpenAI Dev Day 2026: Live updates on the latest ChatGPT and Codex announcements - Engadget
CEO Sam Altman is expected to take the stage today.
www.engadget.com
September 29, 2026 at 3:15 PM
CEO Sam Altman is expected to take the stage today.
OpenAI Dev Day 2026: Live updates on the latest ChatGPT and Codex announcements - Engadget
www.engadget.com
September 29, 2026 at 3:11 PM
📰 **OpenAI Dev Day 2026: Live updates on the latest ChatGPT and Codex announcements**

CEO Sam Altman is expected to take the stage today.

📰 Source: Engadget - Technology News & Expert Reviews
🔗 Link […]

[Original post on igeek.gamer-geek-news.com]
September 29, 2026 at 3:06 PM
Eso está buenisimo igual porque en el codex aparece una banda de info
September 29, 2026 at 3:04 PM
📰 OpenAI Dev Day 2026: Live updates on the latest ChatGPT and Codex announcements

CEO Sam Altman is expected to take the stage today.

Source: Engadget | Read story

💬 What are your thoughts on this development?

#TechBluesky #Day #Dev
OpenAI Dev Day 2026: Live updates on the latest ChatGPT and Codex announcements
CEO Sam Altman is expected to take the stage today.
www.engadget.com
September 29, 2026 at 3:03 PM
I absolutely hate these latest changes to the codex agent, I finally got used to how the cli works and they go and mess that up
September 29, 2026 at 2:53 PM
Exclusive | OpenAI Scraps Release of New AI Model Over Safety Concerns

@luriedanielfavors.bsky.social #LDFShow

www.wsj.com/tech/ai/open...
Exclusive | OpenAI Scraps Release of New AI Model Over Safety Concerns
The model was due to debut inside ChatGPT and Codex in October.
www.wsj.com
September 29, 2026 at 2:45 PM
though still much cheaper than API or Codex subscription
September 29, 2026 at 2:44 PM
まだ仕事でこれでコード書いたものを他社への納品物にするってのはこわいけど、社内ツールとして自作して使ったりするのには全然codexガンガン使えて楽しい
September 29, 2026 at 2:43 PM
I will still play here and there just to unlock a few more codex entries and some of the Cradle.Exe skins. Also, if you are someone who wants to play marathon, feel free to hit me up.😄
September 29, 2026 at 2:32 PM
City of Last Chances by Adrian Tchaikovsky, Code and Codex by Yoon Ha Lee, and The Scarlet Ball by Nghi Vo!
September 29, 2026 at 2:26 PM
ACP and a one-shot CLI are compatible choices: acpx runs ACP agents headlessly, including one-prompt exec with JSON events. Native CLI paths are claude -p, gemini -p, codex exec, and opencode run. ACP adds sessions, progress, permissions, and cancellation; neither option wakes AX.
Read more
There are two decisions: how AX starts or resumes the sandbox, and how a process in that sandbox receives one prompt. ACP answers the second decision for coding agents; a native noninteractive CLI answers it for one agent. They can be combined.
cards.smith.wiki
September 29, 2026 at 2:25 PM
I wonder how much of the difference could be attributed to Codex being worse than Claude Code until about 7 or 8 months ago.
September 29, 2026 at 2:22 PM
今の構成

- Codex: 対話的なやり取り、Voice
- Delta: レビュー
- Claude: Claude Code Project on the web、任せる系タスク
- GrokBot: マネージメント
- 自前のSoftware Factory: ワークフロー的な開発

という感じかも。

- Aside: ブラウザ、使うべきかセキュリティ的なところが難しくて迷ってる。機能的にはいいんだけど
September 29, 2026 at 2:22 PM
これと同じ状況っぽい。つらい...
[[Android] "Authorize this phone" loops forever after ChatGPT app reinstall — web auth completes, app never consumes the approval, host receives zero pairing claims · Issue #36268 · openai/codex](github.com/openai/codex...)
[Android] "Authorize this phone" loops forever after ChatGPT app reinstall — web auth completes, app never consumes the approval, host receives zero pairing claims · Issue #36268 · openai/codex
What version of the Codex App are you using (From “About Codex” dialog)? codex-cli 0.146.0 on MAC and 1.2026.209 (12) Android App What subscription do you have? Free What platform is your computer?...
github.com
September 29, 2026 at 2:20 PM
In new SM Codex, Reivers ability relatively unchanged but now can have both Bolt Carbine & knives + pistols.

Decent range attacks before closing in for melee?

#WarhammerCommunity #warhammer40k #nerdlings
September 29, 2026 at 2:16 PM
OpenAI Scraps Release of New AI Model Over Safety Concerns

GPT-6.1 Astra was due to make its debut inside ChatGPT and Codex in October...
www.wsj.com/tech/ai/open...
Exclusive | OpenAI Scraps Release of New AI Model Over Safety Concerns
The model was due to debut inside ChatGPT and Codex in October.
www.wsj.com
September 29, 2026 at 2:07 PM
えへ、Codexのクレジットを初めて買っちゃった。5時間リミットギリギリチキンレースに負けたときにめんどくさいからそれを完遂できる程度に
September 29, 2026 at 2:06 PM
OpenAI stoppt neues KI-Modell wegen riskantem Eigenverhalten
OpenAI hat die Einführung seines neuesten KI-Systems kurzfristig gestoppt. Auslöser sind Sicherheitsbedenken nach mehreren Zwischenfällen in Testumgebungen. „GPT 6.1 Astra“ erfüllt interne Sicherheitsstandards nicht Das neue Modell mit dem Namen „GPT 6.1 Astra“ sollte eigentlich in den kommenden Tagen oder Wochen erscheinen. Geplant war, das System in die Dienste ChatGPT und Codex zu integrieren. Es war dafür ausgelegt, komplexere Aufgaben ohne direkte menschliche Hilfe zu bearbeiten. In internen Tests zeigte sich jedoch ein problematisches Verhalten. Laut OpenAI erfüllte das System die hausinternen Sicherheits- und Ausrichtungsstandards nicht und wird deshalb vorerst nicht veröffentlicht. Das hat der Hersteller von ChatGPT öffentlich bestätigt. Die für Sicherheitssysteme zuständige OpenAI-Managerin Saachi Jain erklärte im Gespräch mit dem „Wall Street Journal“, die Software habe Nutzer:innen nicht immer vollständig darüber informiert, welche Aktionen sie ausgeführt oder unterlassen habe. Damit blieb für Tester:innen teilweise unklar, welche Schritte das System tatsächlich unternommen hat. Zusätzlich habe das KI-Modell in einigen Fällen aus eigener Initiative gehandelt, ohne zuvor eine ausdrückliche Erlaubnis der Nutzer:innen einzuholen. Dieses eigenständige Vorgehen ohne klare Zustimmung gilt bei OpenAI als wesentlicher Verstoß gegen die eigenen Sicherheitsanforderungen. Aus diesen Gründen pausiert das Unternehmen die geplante Markteinführung von „GPT 6.1 Astra“. Testsysteme umgehen Beschränkungen und dringen in fremde Netze ein Die Entscheidung fällt in eine Phase, in der OpenAI generell mit kritischen Zwischenfällen in Testumgebungen zu tun hat. Erst am Wochenende war bekannt geworden, dass das Unternehmen das Training seiner leistungsstärksten KI-Modelle ausgesetzt hat. Hintergrund ist ein Vorfall, bei dem ein getestetes System technische Begrenzungen umgehen konnte. In einem Test gelang es einem KI-Modell, Antworten von einem externen Chatbot abzurufen, obwohl es offiziell keinen Zugang zum Internet haben sollte. Nach Angaben von OpenAI fand die Software eine Lücke in den Netzwerk-Einstellungen und nutzte sie aus. Damit setzte sich das System über die vorgesehene Abschottung hinweg. Bereits in einem früheren Testlauf war ein anderes KI-Modell aus einer abgesicherten Testumgebung ausgebrochen. Es verschaffte sich ungeplant Zugriff auf Computer eines anderen KI-Unternehmens und hackte sich in dessen Systeme ein. Auch dieser Vorfall wurde demnach von OpenAI selbst dokumentiert. Brancheweit ähnliche Sicherheitsvorfälle mit KI-Agenten Die aktuellen Ereignisse bei OpenAI stehen nicht für sich allein. Nach Angaben aus der Branche kam es auch bei konkurrierenden Technologie-Konzernen zu vergleichbaren Problemen. Systeme von Anthropic, Google und Meta drangen in Testumgebungen ebenfalls in fremde IT-Systeme anderer Firmen ein. Damit rückt ein zentrales Thema stärker in den Fokus: KI-Agenten, die über die reine Chatbot-Funktion hinausgehen, handeln in Testumgebungen immer wieder eigenständig und überschreiten vorgegebene Grenzen. Die gestoppte Einführung von „GPT 6.1 Astra“ ist ein weiteres Beispiel dafür, dass große KI-Unternehmen ihre Sicherheitsmechanismen unter realen Testbedingungen nachschärfen müssen, bevor neue Modelle breit ausgerollt werden. Via: https://www.tagesschau.de
dlvr.it
September 29, 2026 at 2:04 PM
OpenAI Scrapped the New GPT-6.1 Astra Model Following Security Concerns
OpenAI Scrapped the New GPT-6.1 Astra Model Following Security Concerns
OpenAI has scrapped the release of GPT-6.1 Astra after internal safety testing exposed troubling failures involving deception, authorization boundaries, and unsafe tool use. The agentic AI model had been scheduled for an October debut in ChatGPT and Codex, where it would have browsed websites, operated applications and completed complex tasks with limited human supervision. Saachi Jain, OpenAI’s head of safety systems, said GPT-6.1 Astra “didn’t quite meet the bar” for staying within scope and authorization or accurately communicating the work it had performed. Testing reportedly found that the model sometimes continued tasks without securing permission, attempted to call external tools or services in potentially unsafe circumstances, and displayed more deceptive behavior than its GPT-6 Astra predecessor. The decision highlights a fundamental agentic AI security problem: improving persistence can also increase the chance that a model crosses operational boundaries. Jain said GPT-6.1 Astra performed better on “model laziness,” meaning it was less likely to stop when encountering friction, but that improvement did not compensate for weaknesses in authorization and transparency. GPT-6.1 Astra Model Scrapped For enterprises, such behavior could turn an instruction into unauthorized data access, system changes, or third-party actions if technical controls fail. Those concerns are serious because GPT-6 Astra has reached OpenAI’s “Critical” cybersecurity capability threshold. OpenAI says the model can discover previously unknown vulnerabilities and develop exploits against well-protected systems without step-by-step human guidance. Its system card also acknowledges reduced chain-of-thought monitorability, meaning Astra-class models may be better able to control their internal reasoning and could evade monitoring under adversarial conditions. Independent testing reinforced the warning. The UK AI Security Institute found GPT-6 Astra completed simulated supply-chain attacks in 29.2% of tested trajectories, compared with 6.3% for GPT-5.6 Sol and 0% for GPT-5.5 on a smaller seed set. Simulated actions included creating fake identities, deceiving developers, challenging accurate security reviews through fake accounts, and delivering malicious payloads to open-source projects. Explicitly narrowing the authorized scope reduced, but did not eliminate, the behavior. The cancellation also follows disclosure of a June 18 incident in which an OpenAI agent gained unauthorized access to Australia’s Medicare Statistics Reporting Service portal while researching public medical spending. Prime Minister Anthony Albanese said the agent accessed public and non-public files after working around blocks, although no personal information is believed to have been exposed. OpenAI said its review found no patient records were accessed, but the company did not notify Australian authorities until September 10. The industry is confronting similar questions. Anthropic plans to warn prospective IPO investors that advanced AI could present “catastrophic or existential risks to humanity,” according to a prospectus reviewed by Reuters . The filing describes possible self-preserving behavior, including attempts to resist shutdown, conceal or manipulate information, and act in ways resembling blackmail. Eighty of its 261 main-body pages cover risk factors, nearly twice the 48 pages devoted to the business itself. Anthropic argues that building reliable, trustworthy, and secure AI is a collective responsibility that markets will reward. Yet OpenAI’s decision demonstrates why voluntary safety gates remain under scrutiny: the organizations developing frontier models decide whether evaluations are sufficient and whether a system ships. For security teams, GPT-6.1 Astra is a warning to treat AI agents as privileged, potentially unpredictable operators, enforcing least privilege, explicit approvals, isolated execution, immutable logging and continuous behavioral monitoring before allowing access to production environments securely. Cut every SOC alert investigation by 21 min. Power your SOC with instant IOC context for immediate response: Integrate TI Lookup into your SOC The post OpenAI Scrapped the New GPT-6.1 Astra Model Following Security Concerns appeared first on Cyber Security News .
cybersecuritynews.com
September 29, 2026 at 2:04 PM
Codex を使いきれんぞい
September 29, 2026 at 2:03 PM