Eliza Times

Industry News

• OpenAI raises $110B from Amazon, NVIDIA, and SoftBank: Massive funding round values OpenAI at $730B, with strategic partnerships including 3GW of dedicated NVIDIA inference capacity and Amazon collaboration on enterprise products. link

• Block cuts 40% of workforce citing AI: Jack Dorsey attributes 4,000 job cuts directly to AI automation, marking one of the largest AI-driven layoffs to date. link

• Anthropic-Pentagon standoff intensifies: Undersecretary of War calls Anthropic CEO a "liar with a god complex" after negotiations over autonomous weapons and surveillance usage broke down hours before deadline. link

• Sam Altman supports Anthropic's red lines: OpenAI CEO publicly backs Anthropic's stance against autonomous weapons and mass surveillance, revealing OpenAI has same restrictions. link

New Tools & Releases

• Nano Banana 2 (Gemini 3.1 Flash Image): Google's new SOTA image generation model delivers Pro-level quality at Flash speed with enhanced editing capabilities. link

• RepoPrompt 2.0: Fully integrated agent with powerful MCP tools and built-in oracle for seamless codebase navigation and context management. link

• Doc-to-LoRA and Text-to-LoRA from Sakana AI: Meta-learned hypernetworks instantly compile documents into LoRA adapters, improving LLM context windows without training. link

• VGG-T³: Scalable offline feed-forward 3D reconstruction model that finally tackles large-scale multi-view consistency. link

Tips & Techniques

• Don't let agents outpace your understanding: Use plan mode for complex tasks, review code diffs carefully, and ask Claude to explain changes before approving to maintain control of your codebase. link

• How PewDiePie fine-tuned a coding model that beats Llama-4: Built custom GPU rig, generated synthetic data, fine-tuned Qwen2.5-32B and achieved top scores on Aider Polyglot benchmark—demonstrating accessible path to competitive models. link

Research & Papers

• Rubric-based RL overview: Natural evolution from RLVR to non-verifiable domains using LLM judges with granular scoring criteria instead of reward models, with improving results as judge models get better. link

• World modeling needs collaboration, not oneshot generation: Proper immersive worlds require decomposing into problem subsets and iteratively crafting with humans—diffusion video models miss functional interactions. link

• SFT makes models rigid, RL enables exploration: Combining both through reasoning fine-tuning allows models to explore while maintaining hard problem performance. link

• Scaling RL for LLMs guest lecture slides: Rishabh Agarwal shares McGill presentation on the art of scaling reinforcement learning for language models. link

--- *Curated from 500+ tweets across AI research and industry lists*

---

Emerging Trends

✨ OpenAI $110B Funding Round (85 mentions) - NEW Sam Altman announced OpenAI raised $110 billion from Amazon, NVIDIA, and SoftBank, one of the largest funding rounds in AI history. This marks a major escalation in AI infrastructure investment.

✨ Anthropic DoD Dispute (142 mentions) - NEW Major controversy as Anthropic negotiates with Department of War over AI use for surveillance and autonomous weapons. Undersecretary Emil Michael gave detailed Bloomberg interview, with Friday 5pm deadline looming and potential Defense Production Act invocation.

🔥 Claude Code Anniversary & Features (156 mentions) - RISING Claude Code celebrated its one-year anniversary with major feature releases including new demo/recording capabilities, markdown table rendering, and extensive enterprise plugin marketplace with integrations for Google Workspace, Salesforce, Excel/PowerPoint workflows.

🔥 Nano Banana 2 Release (48 mentions) - RISING New image generation model Nano Banana 2 launched with mixed reception. Users comparing quality to Grok Imagine, discussing micro vs macro detail accuracy, and debating performance on specific tests like clock generation.

📊 AI Job Displacement Concerns (118 mentions) - CONTINUING Continued discussion about AI replacing knowledge work, with Jack Dorsey cutting 40% of Block citing AI, Indian IT stocks crashing 6% after Claude Code release, and debates about whether AI will create mass unemployment or simply shift job types.