- Deathmax parentNordVPN calls out when a location is virtual, so unless ipinfo is claiming they have virtual locations that are not labelled as such, they are at least transparent about it. They did document the physical server locations of their virtual locations at launch, but I'm not sure if there's a live doc for new locations. https://nordvpn.com/blog/new-nordvpn-virtual-servers/
- You sort of can on Android, but it's a few steps:
1. Trigger Circle to Search with long holding the home button/bar
2. Select the image
3. Navigate to About this image on the Google search top bar all the way to the right - check if it says "Made by Google AI" - which means it detected the SynthID watermark.
- As if the App Store had any sort of those guarantees. I know of people have been scammed via WebView wrappers that purported to be some benign app to pass app store review, which were then pointed at fake exchange websites afterwards. GitLab which was hosting their C&C mechanism took action faster than Apple or Google did to take down multiple scam apps across multiple different developer identities, but the scammers spun up new apps the next day.
- Riot documents the need to have IOMMU support enabled for Vanguard: https://support-valorant.riotgames.com/hc/en-us/articles/222...
- Vertex's offering of Gemini very much does implicit caching, and has always been the case [1]. The recent addition of applying implicit cache hit discounts also works on Vertex, as long as you don't use the `global` endpoint and hit one of the regional endpoints.
[1]: http://web.archive.org/web/20240517173258/https://cloud.goog..., "By default Google caches a customer's inputs and outputs for Gemini models to accelerate responses to subsequent prompts from the customer. Cached contents are stored for up to 24 hours."
- Gemini uses SentencePiece [1], and the proprietary Gemini models share the same tokenizer vocabulary as Gemma [2, 3, 4].
Out of the large proprietary western AI labs (OpenAI, Anthropic, Google), only Anthropic with Claude 3 and newer lack local tokenizers.
[1] https://github.com/google/sentencepiece
[2] https://github.com/googleapis/python-aiplatform/blob/main/ve...
[3] https://storage.googleapis.com/deepmind-media/gemma/gemma-2-...: "We inherit from the large Gemini vocabulary (256k entries)."
[4] https://storage.googleapis.com/deepmind-media/gemma/Gemma3Re...: "We use the same tokenizer as Gemini 2.0."
- It's a change to the CA rules that was passed in https://cabforum.org/2022/04/06/ballot-csc-13-update-to-subs... to align OV certificate requirements with the EV ones (that enforces the use of HSMs/hardware tokens/etc) that was meant to go into effect for new certificates issued after November 2022, but was delayed and eventually implemented on June 1 2023.
- Since April 2023 they support custom OIDC providers[1], and as of April 2024 that was extended to the free plan as well[2], so you can bring your own auth.
- https://www.minimaxi.com is their website for the Chinese parent company 上海稀宇科技有限公司, https://minimax.io is their international website for the Singapore based company Nanonoble Pte Ltd that handles operations outside of China.
- I did read it, and I even went to their eval repo.
> At the time of writing, there are two major versions available for GPT-4 and GPT-3.5 through OpenAI’s API, one snapshotted in March 2023 and another in June 2023.
openaichat/gpt-3.5-turbo-0301 vs openaichat/gpt-3.5-turbo-0613, openaichat/gpt-4-0314 vs openaichat/gpt-4-0613. Two _distinct_ versions of the model, and not the _same_ model over time like how people like to complain that a model gets "nerfed" over time.
- Your linked article is specifically comparing two different versioned snapshots of a model and not comparing the same model across time.
You've also made the mistake of conflating what's served via API platforms which are meant to be stable, and frontends which have no stability guarantees, and are very much iterated on in terms of the underlying model and system prompts. The GPT-4o sycophancy debacle was only on the specific model that's served via the ChatGPT frontend and never impacted the stable snapshots on the API.
I have never seen any sort of compelling evidence that any of the large labs tinkers with their stable, versioned model releases that are served via their API platforms.
- Gemini's free tier will absolutely use your inputs for training [1], same with Mistral's free tier [2]. Anthropic and OpenAI let's you opt into data collection for discounted prices or free tokens.
- Thoughts used to be available in the Gemini/Vertex APIs when Gemini 2.0 Flash Thinking Experimental was initially introduced [1][2], and subsequently disabled to the public (I assume hidden behind a visibility flag) shortly after DeepSeek R1's release [3] regardless of the `include_thoughts` setting.
At ~10:15AM UTC 04 May, a change was rolled out to the Vertex API (but not the Gemini API) that caused the API to respect the `include_thoughts` setting and return the thoughts. For consumers that don't handle the thoughts correctly and had specified `include_thoughts = true`, the thinking traces then leaked into responses.
[1]: https://googleapis.github.io/python-genai/genai.html#genai.t...
[2]: https://ai.google.dev/api/generate-content#ThinkingConfig
[3]: https://github.com/googleapis/python-genai/blob/157b16b8df40...
- > You can’t just attach an image to your request.
You can? Google limits HTTP requests to 20MB, but both the Gemini API and Vertex AI API support embedded base64-encoded files and public URLs. The Gemini API supports attaching files that are uploaded to their Files API, and the Vertex AI API supports files uploaded to Google Cloud Storage.
- Their primary business model nowadays is as an advertising agency, not book selling: https://www.guinnessworldrecords.com/business-marketing-solu...
- It's simple enough to test the tokenizer to determine the base model in use (DeepSeek V3, or a Llama 3/Qwen 2.5 distill).
Using the text "സ്മാർട്ട്", Qwen 2.5 tokenizes as 10 tokens, Llama 3 as 13, and DeepSeek V3 as 8.
Using DeepSeek's chat frontend, both DeepSeek V3 and R1 returns the following response (SSE events edited for brevity):
which totals to 8, as expected for DeepSeek V3's tokenizer.{"content":"സ","type":"text"},"chunk_token_usage":1 {"content":"്മ","type":"text"},"chunk_token_usage":2 {"content":"ാ","type":"text"},"chunk_token_usage":1 {"content":"ർ","type":"text"},"chunk_token_usage":1 {"content":"ട","type":"text"},"chunk_token_usage":1 {"content":"്ട","type":"text"},"chunk_token_usage":1 {"content":"്","type":"text"},"chunk_token_usage":1