A few new tools for DigIntoAI
I've added some tools to help with the small things that get in the way of using AI. They're still new and a little rough around the edges, so thanks for bearing with me as I improve them.
Follow the changes. Understand what comes next.
At 01:45:00 Asia/Shanghai on November 18, 2026 (OpenAI News RSS; confirmed on @OpenAI at 01:54:03 Asia/Shanghai), OpenAI launched flagship GPT-6.2 Astra (gpt-6.2-astra) and everyday GPT-6.2 Sol (gpt-6.2-sol) on ChatGPT Work, Codex, and the API, bringing Ultrafast to Pro 200 starting with Sol, plus the gpt-6.2-astra-law configuration.
Anthropic's pricing page lists four current Claude models on its API: Claude Fable 5.1 at $10 per million input tokens and $50 per million output tokens, Claude Opus 5.5 at $4 and $20, Claude Sonnet 5.5 at $2 and $10, and Claude Haiku 5.5 from $0.10 and $0.50. This guide covers those prices, the older models still on the price list, and the modifiers that change a bill, including prompt caching, the Batch API, long prompts on Haiku 5.5, US-only inference, fast mode and tool charges. All figures are as of 22:38 Beijing time on October 20, 2026, per Anthropic's pricing page and API docs.
Gemini CLI users who run into a 429 see an error that Google's Gemini CLI FAQ quotes as "API error: 429 - Resource exhausted." Google's documentation says it means a request limit has been exceeded, and it lists what to do next: wait and retry, check your usage, let the CLI fall back to another model, switch to an API key or Vertex AI, upgrade to a higher quota or request an increase. This article covers what the error looks like, what each Google document says it means and the steps those documents give.
When Claude Code can't get past a 529 overloaded_error, it stops with "API Error: Repeated 529 Overloaded errors." Anthropic's documentation says the API is temporarily at capacity, that a 529 is not your usage limit, and that checking the status page, waiting a few minutes or switching models with /model are the things to try. This article covers what the message looks like, how Claude Code retries before showing it and what the docs suggest next.
Since October 18, more than a dozen issues in OpenAI's Codex repository on GitHub have reported the error "Selected model is at capacity. Please try a different model." This article covers what those reports describe and what has worked for the users who filed the reports. OpenAI has not explained the error, and the OpenAI status page shows no incident.
Security firm Claroc says conversations that Muse users share through links can be found by people who were never sent those links. Claroc says it turned up more than 6,800 shared conversations and that some of the pages it opened held personal and business information. According to Ars Technica and Claroc, Meta has not responded.
Vals AI says DeepSeek V4.1 Flash 1017 beat OpenAI's GPT-6 Sol on a multi-step offensive cybersecurity task the evaluation company runs in a simulated corporate network. The results come from Vals AI's own runs, not from either vendor, and Vals AI plans to publish the full evaluation in the coming weeks.
DeepSeek released DeepInfer, the inference framework behind its DeepSeek-V4.1-Flash service, under the Apache-2.0 license, and also open-sourced a token encoding and decoding component called deepseek-recipe. DeepSeek calls DeepInfer "fully open source", but the project README says this release covers offline inference only and lists several pieces a complete serving setup needs that are not included.
Vals AI released Vals Index v2.2 after minor version updates to four component benchmarks. On the leaderboard dated October 19, 2026, Claude Sonnet 5.5 leads, followed by Claude Opus 5.5 and Gemini 4 Argon. Vals AI says DeepSeek-V4.1-Flash-1017 is the highest open-weights model at rank 8. Rankings and scores below come from Vals AI.
Reuters reported, citing people familiar with the matter, that OpenAI is trying to secure investor commitments for a round of about $30 billion before Anthropic's expected mid-November market debut, in a raise that could value OpenAI at over $1.4 trillion. OpenAI and SoftBank declined to comment to Reuters. None of the figures has been announced by OpenAI.
Gemini Notebook’s Android and iOS apps can record up to 1 hour of live audio as a source, per @Gemini_Notebook on October 28, 2026 at 05:49 Asia/Shanghai. Google’s September 15, 2026 blog had previewed the mobile recorder; 9to5Google adds that capture can continue with the screen locked. A September footnote ties related notes-and-sources support to English as the selected output language. The October 19 Outline Review toggle for Slide Decks and Video Overviews is unchanged.
According to the GitHub Changelog, six models were deprecated across most GitHub Copilot experiences on October 19, 2026, including Copilot Chat, inline edits, ask and agent modes, and code completions. GitHub suggests GPT-6.1 Sol, GPT-6 Luna, Gemini 3.8 Flash and Grok 4.7 as replacements. On the same date, GitHub made its Copilot model policy REST API generally available.
In a blog post dated October 19, OpenAI says its latest models are stronger at building and migrating code in low-resource languages such as COBOL, Fortran and Visual Basic. On OpenAI's MultiLangBench, GPT-6 Astra reaches an 87.4% average test pass rate, versus 26.8% for GPT-5.6 Sol. On OpenAI's COBOL migration evaluation, GPT-6.1 Sol reaches 86.6% at low reasoning effort and 97.9% at higher effort.
Personal experiences, thoughts and notes.
I've added some tools to help with the small things that get in the way of using AI. They're still new and a little rough around the edges, so thanks for bearing with me as I improve them.
There’s a new place on DigIntoAI to find free APIs, signup credits and limited-time offers. If you want to try a model without paying upfront, have a look.
Tibo posted on X that Day 8's reset will land globally by the end of the Pacific time workday.