TLDR AI 2026-09-25

As AI agents take on more work, cloud inference costs can grow with every task. The AMD Tokenomics Calculator helps you understand the economics of your AI workloads and optimize token spend.

• Understand your AI costs. Model infrastructure costs to see how token usage translates into ongoing spend.
• Unlock the economics of local AI. See how running AI locally can reduce cloud inference costs and decrease token spend.
• Reduce recurring AI costs. High-performance agentic PCs featuring AMD processors run multiple agents locally, completing tasks up to 6x faster while reducing reliance on cloud inference.

👉 Calculate your potential savings now

Meta has introduced Muse Realtime Avatar, a technology that transforms conversations into expressive, interactive avatars in real time. Using a seamless integration of voice and visual performance, it synchronizes speech and avatar expressions accurately in live interactions. Muse Realtime Avatar outperforms competitors with improved visual quality and real-time responsiveness, supported by Meta's optimized AI inference stack.

Google launched Gemini 3.8 Live featuring a Live Avatar, enhancing user interaction. This update offers real-time responses and more personalized user experiences. It positions Google to compete more aggressively with other AI platforms.

OpenAI appears to be preparing a new ChatGPT Pro Max subscription priced at $500 per month. It is unknown whether the company plans to justify the $500 price through speed, larger usage allowances, longer-running Work sessions, or a combination of the three. The timing of the leak is notable as OpenAI's DevDay will take place on September 29. The company is expected to discuss APIs, developer tooling, new subscription tiers, and new products at the event.

1 in 4 AI agents still deploy without runtime monitoring, per New Relic's Observability Forecast survey of 2.5k IT and engineering pros. And as organizations scramble to close the gap, tool sprawl is creeping back in. To see what sets leaders apart from those playing catch-up, browse the report

Contrastive Language Models (CLMs) are a new class of System One model trained with a contrastive learning objective that connects states and actions. CLM-8B delivers performance comparable to Jev across computer-use, gaming, and tool-calling tasks while achieving up to 9x lower latency. It also sets a new state-of-the-art on changing agentic coding benchmarks. The model is pre-trained on 60M Nemotron Q&A pairs, mid-trained on 30M synthetic hard negatives, and post-trained on 1M agentic trajectories.

Managed Deep Agents combines the Deep Agents harness with all of the infrastructure required to run agents in production. It is the simplest way to build, deploy, and run mission-critical agents in production. Managed Deep Agents 0.8 adds support for user-owned credentials, user-level memory, HTTP channels, and file transfer in Slack. It also adds a pre-built tool for web search powered by Parallel.

Recent frontier models are rapidly saturating old benchmarks and moving into harder real-world, scientific, and agentic tasks. The pace of capability gains, alongside soaring AI usage and revenue, is shifting debate from workplace augmentation toward recursive self-improvement and superintelligence.

Hugging Face says its autonomous-agent cyberattack exposed three priorities: stronger incident transparency, reducing capability asymmetries between attackers and defenders, and preserving open-source access for defense. AI created new attack risks but also helped investigate, mitigate, and harden systems.

TLDR is hiring its first PM to help build the agent-first operating layer used across the company. We're looking for a builder who has shipped real products/systems with LLMs. Click here to learn more.

DAYJOB: Finance tests whether AI agents can complete 80 realistic finance assignments using supporting documents and produce analysis a professional can use.

After five-minute interviews, Claude agents traded books for employees, and their preference rankings matched the humans' on 61% of pairs.

Taste-Bench evaluates whether an LLM agent can choose the better next step at consequential forks in long-horizon tasks.

Trajectory.ai focuses on "intelligence density," measuring cost per task instead of cost per token for more efficient AI model training.