- Lore Brief
- Posts
- Issue #164: America Launches a Tech Force
Issue #164: America Launches a Tech Force
PLUS: Gemini 3 Flash goes instant, space compute finally clicks, and frontier AI keeps accelerating
Nathan Lands
December 19, 2025
Good morning, welcome to this week’s Lore Brief, your 3-minute brief of the most important moves in AI and tech.
This issue is brought to you by Factory, the best agentic development platform. We used it to create the new Lore.com website!
Google rolls out Gemini 3 Flash → The official thread announces a free, unlimited, and lightning-fast model upgrade optimized for everyday tasks. It outperforms prior frontier models like Gemini 2.5 Pro on many benchmarks. See more here →
OpenAI launches ChatGPT Images with new flagship model → The update powers native generation and precise edits in ChatGPT, with better instruction following, detail preservation across changes, and up to 4x faster outputs. A new Images tab adds presets, trending prompts, and one-time likeness uploads for quick creative sessions. Read more here →
NVIDIA unveils Nemotron 3 open models family → The efficient MoE series spans Nano (30B, out now with 1M context), Super (100B), and Ultra (500B) for reasoning and agents. All include datasets and RL tools; Nano leads open efficiency benchmarks. Read more here →
The U.S. Government Just Announced Its Own Elite Tech Force
This Visualization Shows Why Space Compute Is a Big Deal
Google Just Made Frontier-Level AI Fast Enough to Feel Instant
Meta releases SAM Audio for advanced sound separation → This open-source unified model isolates specific sounds from complex audio using text, visual, or time-span prompts. It tops benchmarks and ships with datasets to enable new editing and perception apps. Read more here →
Google boosts Translate with Gemini upgrades → Text translations now better capture idioms and slang in 20+ languages; beta speech-to-speech preserves tone across 70 languages. New practice tools add feedback and streaks for learning. Read more here →
Black Forest Labs drops FLUX.2 [max] → The pro-tier model adds real-time web search for grounded facts like weather or scores, handles up to 10 reference images for consistency, and ranks #2 on key image benchmarks. It excels at polished product visuals and cinematic creative work.
Read more here →Zoom AI sets new record on Humanity's Last Exam → Federated system scores 48.1% on the full benchmark, edging past prior SOTA with agentic explore-verify workflows. It blends small and frontier models for scalable enterprise tasks. Read more here →
Glif shares agent for automated renovation videos → Start with one image; the agent generates staged changes, adds creative Kling transitions, and compiles the final clip. Everything runs inside without manual editing.
See more here →