pplx-decider-27b open-sourced; Decisions API priced at 4¢ per million input tokens

#1 𝕏

Aravind Srinivas says pplx-decider-27b, a multimodal Decision Model, is being open sourced and offered through a new Decisions API at 4 cents per million input tokens with free output tokens. He says the price is intended to drop further over the coming days.

Also covered by: @Harrison Chase

#2 𝕏

Boris Cherny announced Claude mods, which let users customize how Claude works and looks through prompting. Mods can also be shared as plugins for others to try.

Also covered by: @Thariq, @Thariq

#3 𝕏

Mustafa Suleyman announced the launch of an unnamed real-time transcription model, claiming it ranks #1 globally and is 55% faster and 60% cheaper than ElevenLabs. He invited developers to build agents on an unspecified platform.

#4 𝕏

Google DeepMind announced SynthID Bio, a new family of watermarking methods for AI-generated biological designs. The company describes it as a world first, embedding imperceptible signatures directly into protein sequences without affecting their biological function.

Also covered by: @Google DeepMind

#5 𝕏

Josh Woodward announced the Stitch CLI as a way to get design ideas on demand.

#6 𝕏

Ali Ghodsi said ai_decide() had just been released, running decision models natively on data in Databricks. He said it enables system one decisions on big data, framing fast typed decisions as a game changer compared with slow text generation.

Also covered by: @How I AI Podcast

#7 𝕏

Sam Altman commented that 6.1 Sol was their fastest-growing model ever and had been somewhat slow under load, but should perform much better now.

#8 𝕏

Cursor announced that GLM 5.3 and GLM 5.3 Flash are available in Cursor, while GLM 5.3 Max is the best-scoring open-weight model on CursorBench 4.0.

Also covered by: @bolt.new

#9 𝕏

The post shares a broadcast about building visual AI agents from a prompt using NVIDIA Cosmos and VSS 3.3.

#10 ▶️

The one OpenAI announcement that can actually make you money...

Fireship

Also covered by: @Dan Shipper

#11 𝕏

Garry Tan highlighted Modus’s open-source test designed to assess whether AI can perform actual junior-auditor work, rather than merely answer accounting questions. He predicted that every job will eventually have similar evaluation benchmarks.

#12 𝕏

Teresa Torres recapped how Vistaly rebuilt its entire product to explore AI-powered opportunity solution trees, noting that the agents were the easy part. The episode covers interview synthesis, evaluations, guardrails, orchestration changes, data residency on Bedrock, change sets, and product-manager discovery workflows.

#13 𝕏

Harrison Chase shared a four-step approach to model routing: understand the tasks, understand the models, build the router in the harness, and track outcomes. The goal is to lower costs without a performance hit.

#14 𝕏

Hugging Face announced MCP support in Hugging Face Chat’s ML Intern mode, enabling users to bring their data into the linked chat experience.

#15 📝 OpenAI News

How Albertsons Companies is reimagining retail from the inside out - OpenAI News announced on October 1, 2026, that Albertsons Companies is expanding its work with OpenAI to improve the retail experience for millions of customers.

#16 ▶️

Masterclass: How FDEs make $1M/yr deploying AI agents

Greg Isenberg

#17 𝕏

The post shared instructions for building a Bolt.new site that adapts based on UTM parameters.

#18 𝕏

A library of Amex-curated, ready-to-use skills was launched in Perplexity Computer for eligible U.S. Amex Business Small Business Card Members. The pre-built workflows support tasks such as forecasting cash flow and generating marketing campaigns without requiring users to write instructions from scratch.

#19 𝕏

Santiago says Nace AI released Drex 1.5, a decision model with a 128k context window that scores choices across entire contracts or video transcripts without chunking. The under-10B-parameter model costs $0.04 per 1M input tokens and ranks #1 on Decision Index 0.2.1 with 58.28 versus Jev 1.13.0’s 57.91. Nace AI is also offering 250M free tokens to the first 10,000 builders.

#20 𝕏

Guillermo Rauch commented that tokens aren’t served when ZDR requirements are set and no suitable providers are available, while noting there are plenty of fallbacks and capacity.