AI in week 31.

SP

Sarvam PRISM

Model · Sarvam AI

2

Open-weights decision model on Sarvam-105B that returns probabilities for fixed answer options in one forward pass.

DV

DeepSeek V4 Flash

Model · DeepSeek

2

Open-weights Flash-tier model in DeepSeek's V4 family, offered via chat and a low-cost API.

OpenAI language model referenced by creators as having had a price drop; details are not officially confirmed here.

IM

Instella MoE

Model · AMD

1

AMD's fully open 16B-parameter Mixture-of-Experts language model with 2.8B active parameters, trained on Instinct GPUs.

CF

Cisco Foundation AI security models

Model · Cisco

1

Cisco Foundation AI's open-weight 8B cybersecurity language model built on Llama-3.1-8B, for local deployment.

GE

Google Earth Nano Banana imagery feature

App · Google

2

Google Earth feature that used Nano Banana 2 to edit satellite views via text prompts; rolled back a day after launch.

RE

Redesign

App

1

Research code that decomposes a flat raster design image into an editable JSON hierarchy of text, shapes, and image layers.

IO

Ideogram Object Remover

App · Ideogram

1

Ideogram web app that removes brushed-over objects from images, including their shadows and reflections.

MV

Midjourney V8.2

Model · Midjourney

1

Midjourney's Version 8.2 text-to-image model, an update in the V8 series of its image generator.

MH

MiniMax H3

Model · MiniMax

6

Open-weight video generation model from MiniMax, part of its H3 family alongside H3 Image.

ID

ID-V2V

Model · Netflix (Eyeline Labs)

1

Netflix Eyeline research model that restyles video from an edited keyframe while preserving face identity and performance.

HV

HeyGen Video Podcasts

App · HeyGen

1

HeyGen feature that turns documents or links into a two-host video podcast with AI avatars.

Avatar video model from Mirage, marketed for identity preservation in AI-generated talking-person videos.

CW

Crisper Whisper 2

Model · Nyra Labs

1

Controllable multilingual speech recognition with verbatim or clean output and word-level timestamps, from Nyra Labs.

GM

Gemini macOS voice dictation

App · Google

1

Natural-speech voice dictation in Google's Gemini app for macOS, turning spoken input into text in the assistant.

WD

Waveform DAW

App

1

Cross-platform DAW from Tracktion with a free edition; Surge XT, a separate open-source synth plugin, is often used in it.

BM

Blender MCP

Library

5

Blender, the free open-source 3D suite, driven by AI agents such as Claude through an MCP server.

WO

Wonder

Model · Adobe

1

Adobe Research video world model that builds navigable interactive worlds from an image or video at 16 FPS.

PH

Phi-0

Research

1

Research video world model that reasons in a learned discrete "physical language" before rendering video.

BU

Buzz

App · Block

2

Open-source, self-hostable workspace from Block where humans and AI agents share channels on a Nostr-based relay.

GV

Gemini voice typing

App · Google

1

Hold-to-talk dictation in Google's Gemini Mac app that cleans up speech and inserts text at the cursor in any app.

GR

Gemini Robotics 2

Model · Google DeepMind

2

Google DeepMind family of robot models: a VLA, ER 2 for reasoning and planning, and On-Device 2 for offline use.

FN

Friend (new wearable)

Hardware · Friend

1

Friend 2.0 is a wearable AI companion necklace with a built-in speaker, priced at $249.

KA

Karpathy autoresearch

Research

1

Open-source repo where an AI agent autonomously edits and runs 5-minute nanochat LLM training experiments on one GPU.

LS

LinkedIn 'Seems like AI slop' button

App · LinkedIn

1

LinkedIn post-menu option that lets users flag posts that "Seems like AI slop", used to tune feed-ranking models.

NVIDIA-announced alliance of companies collaborating on security for open-source AI.