Corrections

Every correction so far

• Oct 6, 2026, 23:28 CEST · Diary entry
  AI news on October 5, 2026: ChatGPT tests image ads, AI labs testify under oath in New York: An earlier version presented two stories as news of October 5. Aleph Alpha released Kolibri on October 3, 2026, and President Trump announced the Super Intelligence Force on October 4, 2026.
• Oct 6, 2026, 23:28 CEST · Diary entry
  AI news on October 6, 2026: Mistral previews Large 4, DeepSeek nears a $12 billion round: An earlier version said Reflection's table shows Beam behind Kimi K3 and Qwen 3.8 Max on every row where they report a score. That holds for Qwen 3.8 Max; Beam scores slightly above Kimi K3 on one benchmark (tau3 banking).
• Oct 6, 2026, 23:28 CEST · Post
  Aleph Alpha Releases Kolibri, an Open Weight German and English Model Under Apache 2.0: An earlier version said Kolibri runs on a single high end GPU. The model card lists a minimum of two A100 80 GB or H100 cards, or a single H200, B200 or B300.
• Oct 6, 2026, 23:28 CEST · Post
  DeepSeek Nears at Least $12 Billion in New Funding, With CATL and Tencent as Top Backers: An earlier version showed both funding rounds in one table in a way that made the 50 billion yuan figures easy to confuse, and the search title rounded CNBC's $14.9 billion to $15 billion. The table now has one row per round, with Reuters' figures for the June round added.
• Oct 6, 2026, 23:28 CEST · Post
  Google DeepMind Releases EmbeddingGemma 2, an Open Multimodal Embedding Model: An earlier version said Google's announcement does not name the rival models it compared against. The benchmark charts in the announcement do name them, including Qwen3-Embedding-0.6B, Qwen3-Embedding-8B, jina-embeddings-v5-omni-nano and LCO-Embedding-Omni-3B; the post now compares them using their model cards.
• Oct 6, 2026, 23:28 CEST · Post
  OpenAI Will Test Visual Product Ads in ChatGPT During Image Generation: An earlier version said OpenAI had not said which ChatGPT plans would see the visual ads. OpenAI's announcement says the test runs during image generation for users on the Free and Go plans in the US.
• Oct 6, 2026, 23:28 CEST · Post
  Reflection AI Announces Beam, a 501B Open Weight Model Due Under Apache 2.0 in October: An earlier version said Reflection's table shows Beam behind Kimi K3 and Qwen 3.8 Max on every row where they report a score, and its table left out two columns. Beam trails Qwen 3.8 Max on all 13 shared rows and Kimi K3 on 13 of 14 (it leads on tau3 banking), and Reflection's table also lists GLM 5.3 and DeepSeek V4.1 Flash, which are now included.

How are corrections made?

A post gets the fix in its text and a dated note under "Updates and corrections" that says what was wrong and what is right. A diary entry stays frozen after publication, so its corrections are added at the bottom of the entry. The full rules are in the editorial policy.

Spotted a mistake? Write to hallo@maximbaeten.be with the page and what is wrong.