AI News
+4
Aug 23, 2026
•
4 min read
Black Forest Labs released a video upscaler on 20 August. You can now make many low cost takes and pay for quality once, at the end. One setting is on by default, and it quietly changes faces.
Suno
Aug 22, 2026
Suno put MIDI on the Studio timeline, and a MIDI clip can now be the prompt for a new generation. The feature is useful. The sentence Suno wrote underneath it is more useful, because it says out loud where describing music in words stops working.
Aug 18, 2026
The app and the website went dark on the twenty-sixth of April. Everything that is left switches off on the twenty-fourth of September. On the table where OpenAI lists what to move to, the row for Sora holds a dash.
AI Images
Aug 17, 2026
Google switched off three Imagen 4 endpoints today. What replaces them has no separate image call at all: pictures now come out of the same request as words, and that request will accept fourteen reference images, a video, or a link to one.
Aug 16, 2026
Google shipped a workhorse model on the thirteenth of August and put an introductory number on it: three quarters of a dollar per million tokens in. That number holds until the thirty-first of December. On the first of January it doubles.
Aug 10, 2026
Meta published a 30 billion parameter multimodal model today, small enough for one consumer card. The size will get the headlines. The three words that matter for paid work are Apache two point zero.
+3
Aug 7, 2026
Liquid AI put a 2.6 billion parameter model on Hugging Face on August 4 that runs in under 2.5 GB of memory. It makes no pictures at all, which is exactly why it is worth your attention this week.
Multimodal AI
Jul 28, 2026
FLUX 3 is in early access, not general release. Its larger idea is a useful creator habit: write image, motion, sound, and action as one coherent world.
Prompt Craft
+2
Jul 25, 2026
3 min read
Leaderboards rank models in general, which is a question nobody's workflow actually asks. A bench ranks them by the job. The five seats in our current rotation, and what each model must keep doing to hold its place.
Evaluation
Jul 24, 2026
Google's first ATLAS report studied 15 million interactions and found that fewer than one in ten workplace uses fully automate a task. The useful pattern is collaboration with a named handoff.
Jul 23, 2026
Google's new Flash pair turns model choice into a routing decision: one lane for difficult planning and multimodal judgment, another for repeated extraction, classification, and subagent work.
