APIDir Signals

Know what changed, who it affects, and what to do next.

A source-linked change graph for AI companies, products, model families, exact models, APIs, prices, tutorials, and real external discussions.

Not another AI news riverOfficial evidence firstOne event can belong to many entitiesAction required is separate from popularityCommunity discussion never substitutes for fact

Tracked hubs

Switch by company, product, model, or capability

View every entity

Change intelligence

30 source-linked results

Reset
Deprecation Importance 92/100 Confidence 100/100

Codex MCP server command deprecated

OpenAI deprecated the Codex MCP server command. Integrations that launch it should move to the Codex app server.

Action requiredReplace Codex MCP server launch commands with the Codex app server before the old path is removed.
Deprecation Importance 88/100 Confidence 100/100

DeepSeek-V4 ​

DeepSeek-V4 ​ The DeepSeek API now supports V4-Pro and V4-Flash, available via both the OpenAI ChatCompletions interface and the Anthropic interface. To access the new models, the base_url remains unchanged, and the model parameter should be set to deepseek-v4-pro or deepseek-v4-flash . The two legacy API model names, deepseek-chat and deepseek-reasoner , will be discontinued in three months (2026-07-24). During the current period, these two model names point to the non-thinking mode and thinking mode of deepseek-v4-flash , respectively. For more details, please refer to this documentation .

Action requiredReview affected integrations and migrate before the documented retirement boundary.
Pricing Importance 82/100 Confidence 100/100

Discounted background completion windows in the Responses API

Nov 17, 2026 API Beta Share Discounted background completion windows in the Responses API Choose a two-hour or eight-hour background completion window in the Responses API public beta. Model token rates are 30% lower than Standard for the two-hour window and 40% lower for the eight-hour window. View source (opens in a new window)

Pricing Importance 82/100 Confidence 100/100

OpenAI CSAM Image Classifier

Nov 16, 2026 API GA Share OpenAI CSAM Image Classifier Approved organizations can use the OpenAI CSAM Image Classifier at no cost to help prioritize images for human review. The model returns separate signals for potential child sexual abuse material (CSAM) and potential AI-generated CSAM. View source (opens in a new window)

Pricing Importance 82/100 Confidence 100/100

ChatGPT for Healthcare through AWS Marketplace

Nov 16, 2026 ChatGPT GA Share ChatGPT for Healthcare through AWS Marketplace Eligible healthcare organizations can now purchase ChatGPT for Healthcare through AWS Marketplace using their existing AWS billing relationship. This provides a purchasing option for organizations that want to deploy ChatGPT for Healthcare across clinical, research, and administrative teams. Contact your OpenAI or AWS account team to confirm eligibility. Your workspace’s feature availability, admin controls, and applicable agreement requirements still apply. View source (opens in a new window) Help center (opens in a new window)

Pricing Importance 82/100 Confidence 100/100

Introducing ChatGPT Wallet

Nov 13, 2026 ChatGPT GA Share Introducing ChatGPT Wallet ChatGPT Wallet helps your dot go from finding what you need to making a purchase on your behalf. Whether you’re shopping for a new sweater or booking a hotel, you can ask your dot to help and use Wallet when it’s time to check out. Save an eligible card once, then approve purchases without entering your payment details again. Before your dot pays, you review the merchant, items, total, and selected card, then authorize the purchase using biometrics such as Face ID or Touch ID. Wallet provides a substitute payment credential for checkout, keeping your actual card details hidden from the agent and business. You retain the rewards associated with your card and its protections for qualifying disputes. Wallet is rolling out gradually in dots to ChatGPT Pro subscribers age 18 or older in the United States, with an eligible US-issued card and US billing address. To get started, ask your dot to add a card to Wallet. Some websites may not support purchases made by agents, and approval alone doesn’t confirm an order—check the merchant’s confirmation after checkout. View source (opens in a new window) Help center (opens in a new window)

API change Importance 78/100 Confidence 100/100

DeepSeek-V4-Pro Update ​

DeepSeek-V4-Pro Update ​ The GA release of DeepSeek-V4-Pro has been rolled out on the APP, Web, and API. The API calling method remains unchanged — simply set the model name to deepseek-v4-pro to use the latest version. Significantly enhanced Agent capabilities The GA version of DeepSeek V4 Pro greatly enhances agent capabilities, with particularly significant performance improvements in production environments. HLE (wo / w tools): 42.7/60.0 Terminal Bench 2.1: 87.9 NL2Repo: 61.5 Cybergym: 83.3 DeepSWE: 62.7 Toolathlon-Verified: 74.1 Agents' Last Exam: 25.7 AutomationBench (Public): 31.8 DSBench-FullStack: 71.1 DSBench-Hard: 67.2 Native support for the Responses API The DeepSeek API now natively supports the OpenAI Responses API format and is specifically adapted for Codex. Users can refer to the official documentation and complete the Codex configuration with a one-click configuration script. More flexible thinking effort control The thinking modes of V4-Pro and V4-Flash now support three thinking effort levels: low / high / max. In real-world usage, users can flexibly choose based on task complexity: use low for simple tasks, high for daily Agent tasks, and max for more complex s

Release Importance 76/100 Confidence 100/100

Clinical search and healthcare plugins

Nov 16, 2026 ChatGPT GA Share Clinical search and healthcare plugins We’ve improved clinical search in ChatGPT for Clinicians, with updates to how ChatGPT finds, evaluates, and cites medical evidence. Healthcare plugins also let you connect supported tools to ChatGPT. Plugin availability and requirements vary. We’ve improved clinical search in ChatGPT for Healthcare and expanded the selection of healthcare plugins, helping teams bring clinical evidence and connected tools into their conversations. Available plugins depend on your workspace settings and provider requirements. View source (opens in a new window) Help center (opens in a new window) Help center (opens in a new window)

Community Importance 32/100 Confidence 55/100

Show HN: Terminess, a harness that runs the official Claude Code and Codex CLIs

TLDR: I built a terminal (harness) for Claude Code and Codex that works the official way, on your own subs. Anthropic stopped Claude subs working in 3P harnesses in Feb, 9 months of suffering later, I decided to build one that plays by their rules, a terminal (harness) that runs Claude Code and Codex (and others), together. (I am not claiming that this is a Deepseek harness killer lol, but it works really well for me) I use a number of agents in tandem, on my laptop and my server. Checking them from my phone took 4+ apps and I was losing hours away from my laptop, and copying context between them ate the rest. Terminess puts it all in one place. It complies with all their terms. Terminess runs the real claude and codex CLIs, unmodified, signed in with your own plans, exactly as your reg mac terminal or iTerm would. It doesn’t read, store or reuse your login tokens, and never proxies your requests, so nothing about it works around Anthropic's or OpenAI's terms (currently). Underneath is its own terminal engine, written from scratch for agent work. - Sessions live on your Mac or your own server, so agents keep going with the lid shut - You can start on your phone and pick it up on yo

Community Importance 32/100 Confidence 55/100

Show HN: OpenButterfly – Tracking the environmental impact of the AI ecosystem

I built OpenButterfly to help people understand the environmental impact of the AI ecosystem. Openbutterfly tracks major AI companies (AI labs, cloud providers, chip makers and data center operators) and maps data center sites with their local grid carbon and water stress. Every disclosed figure links to the report page it came from. Missing figures show as "not disclosed", and estimates are clearly labeled. A few things surprised me. Eight of the 35 companies publish no energy, emissions or water figures at all, and seven of those eight are AI developers, including OpenAI, Anthropic, xAI and DeepSeek. No company publishes how much electricity its AI workloads use. And because of renewable energy certificates, companies typically report only about a fifth of the emissions from the grid power they actually used. Openbutterfly website itself is maintained by local LLMs (Gemma 4 31B, Qwen 3.8 27B), so its own AI environment footprint stays small. Automated pipelines running local LLMs fetch, extract company figures, and regularly they pick relevant stories from publisher news feeds and write a summary of each.

Community Importance 32/100 Confidence 55/100

Show HN: Feedback/AMA MeditionBench; LLMs for mediation and conflict resolution

I am the founder of Human Assisted Intelligence and I believe AI can be used to reduce polarization and improve the way people communicate and reach agreement. At https://mediationbench.com/ , there are many frontier LLMs tested, but it is difficult to keep up. I've been trying to get more free openrouter or huggingface credits. Azure for OpenAI and Anthropic have both provided token credits. (Thank You!) The research paper is there as well. There have been efforts to use LLMs to lie and play diplomacy, which are excellent compliments to this work. In v4, the test will get harder as Claude nearly saturated the test. Astra was surprisingly low, perhaps because of alignment or safety efforts. I'd love to get feedback and hear from all of you. I've been working on related ideas for 20 years, so it's definitely a passion project with too much time in isolation.

Community Importance 32/100 Confidence 55/100

Ask HN: What models do you use for tasks Anthropic/OpenAI don't approve of?

People frequently talk about the ever growing lists of tasks that claude and friends will "refuse" to work on, and then frequently someone will make a comment about "open models" or even specifically uncensored/"abliterated" or something, but are rarely specific. If you just search you can find a lot of models claiming some level of "uncensored" but presumably you also need them to be some level of capable. So, for those of you who need to do "unapproved" things, what are your go-tos?

Community Importance 32/100 Confidence 55/100

Ask HN: Why doesn't Ramp AI Index include OpenRouter or other aggregators?

I spend a lot of time looking at these two open data resources, which I'm sure many of you do as well: Open Router Rankings, especially the first raw tokens per week by model (first chart) https://openrouter.ai/rankings And also Ramp AI Index token volume by model maker https://ramp.com/data/ai-index?metric=token-volume&detail=maker&mode=volume What I can never square with these two charts is how they affect each other? One of my biggest questions, is why does Ramp only really show Anthropic/OpenAI/X? There must be hundreds of their companies spending large amounts on OpenRouter, but this doesn't seem to show up in any of their reports. Perhaps since OpenRouter is an aggregator of models, Ramp can't figure out how $1 of spend is used, Ramp can't see which model it was spent on? But still, they could at least have included OpenRouter and others as an "aggregator" or something to give an idea of the marketshare and growth? Does anyone know any public datasets that help merge these two? Or have a better way of understanding how they fit together?

Community Importance 32/100 Confidence 55/100

Why Google, Microsoft, Anthropic and OpenAI need their own AI detection tools

I'm not naive enough to think that these companies don't already have some internal detection tools, but they clearly are not being used enough. Lately researching anything using agents has been a pain (especially if the query leads up to buying intent) - the sources quoted are often AI slop, which are then used to deterministically argue a point. On the first challenge to logic, the reply is usually an apology followed by a tangential search and agreement on the user argued point - sometimes even a hedge. There is a simple problem embedded at the center of this, that at least partially explains this degradation - AI generated text. There is now arguably more AI generated text on the web than human writings, and platforms that used to be very credible (attracting high quality technical writers) now seem to be full of AI generated slop (someone has to call them out). It's a pattern I'm sure most people are noticing under the hood - query -> beautiful prose -> suspect sources. You challenge the answer, and all of a sudden it's a wildly different conclusion. To me that just sounds unreliable, because you could potentially argue with it to generate any conclusion you want - doesn't sou

Publication boundary

Confidence, importance, and discussion are three different numbers.

Official and first-party repository changes may publish automatically after deterministic validation. Independent sources remain draft until review. Community links are attached as discussion evidence and cannot create an indexable fact by themselves.

Read the evidence method

Build your watchlist

Follow the systems that can break your product or budget.

Save events and hubs locally today. Confirm your email for APIDir research updates, or request team access for entity watchlists, migration alerts, API exports, and webhooks.

See Signals plans