Poolside Launches Laguna XS.2 as Open Source AI Models Gain Traction

Poolside Releases Free Laguna XS.2 Model for Agentic Coding

American AI startup Poolside on Tuesday launched its Laguna XS.2 open source model, targeting local agentic coding workflows as the company seeks to compete with proprietary models from OpenAI and Anthropic. According to VentureBeat, the San Francisco-based company founded in 2023 released two new Laguna large language models optimized for AI agents that can write code, use third-party tools, and take autonomous actions.

https://x.com/eisokant/status/2049142230397370537

Poolside also introduced “pool,” a coding agent harness, and “shimmer,” a web-based mobile-optimized development environment for interactive coding previews. The release comes as Chinese companies like DeepSeek and Xiaomi have gained attention by offering near-frontier AI capabilities at significantly lower costs than their American counterparts.

Chinese AI Companies Drive Open Source Innovation

SenseTime, the sanctioned Chinese AI firm, released its SenseNova U1 image model on Tuesday, claiming faster image generation and interpretation than top US competitors. According to Wired, the model can process images directly without first translating them to text, reducing computing power requirements and speeding up operations.

“The model’s entire reasoning process is no longer limited to text. It can reason with images as well,” said Dahua Lin, SenseTime’s cofounder and chief scientist. The company released U1 for free on Hugging Face and GitHub, continuing the trend of Chinese firms becoming major contributors to open source AI development.

Ten Chinese chip designers, including Cambricon and Biren Technology, announced hardware support for U1 on release day. This compatibility matters given US export controls that restrict Chinese access to advanced AI chips, particularly those from Nvidia used for training.

Image Models Drive Mobile App Growth

Image AI model releases now generate 6.5x more mobile app downloads than traditional conversational model updates, according to new data from app intelligence provider Appfigures. TechCrunch reported that this marks a significant shift from earlier periods when conversational AI improvements drove user adoption.

Google’s Gemini added 22+ million downloads in the 28 days following its Gemini 2.5 Flash image model release last August, lifting app downloads by more than 4x. ChatGPT generated over 12 million incremental installs after introducing its GPT-4o image model in March 2025 — roughly 4.5x more downloads than its GPT-4o, GPT-4.5, and GPT-5 text model releases combined.

Meta AI’s video feed feature “Vibes” contributed an estimated 2.6 million incremental downloads in the 28 days after its September 2025 release. However, Appfigures cautioned that increased downloads don’t always translate to higher mobile revenue.

Enterprise AI Security Concerns Rise

Cisco on Thursday released its open source Model Provenance Kit to help organizations track and verify third-party AI models. SecurityWeek reported that the tool addresses security, compliance, and liability issues stemming from unverified claims by model developers about their models’ sources, vulnerabilities, and training biases.

“If unaccounted for, those vulnerabilities can continue to propagate, whether they affect an internal chatbot, an agent application, or a customer-facing tool,” Cisco explained. Organizations often lack visibility into changes made to models from repositories like HuggingFace, where millions of models are available with varying levels of documentation quality.

The tool aims to help enterprises avoid deploying poisoned or manipulated models while enabling better incident response and remediation through improved model lineage tracking. Cisco highlighted risks including licensing violations and regulatory compliance issues related to government AI documentation requirements.

Meta Shifts Strategy with Proprietary Muse Spark

Meta introduced its new AI model Muse Spark at the beginning of Q2 2026, marking a departure from the company’s previous strategy of releasing Llama models as free open source software. CNBC reported that investors are watching CEO Mark Zuckerberg’s commentary about the company’s AI strategy during upcoming earnings calls.

Analysts at Citizens described AI as a “complementary good” for Meta’s core social media business. The shift toward proprietary models represents a significant strategic change for the company, which had previously championed open source AI development through its Llama model series.

The timing of Muse Spark’s release positions Meta to compete more directly with proprietary offerings from OpenAI and Anthropic, though specific performance benchmarks and pricing details remain undisclosed.

What This Means

The AI model landscape is fragmenting along geographic and business model lines. US companies are increasingly caught between proprietary high-performance models and the need to compete with free, high-quality alternatives from Chinese firms and smaller American startups.

The surge in image model adoption suggests users value multimodal capabilities over pure text performance improvements. This trend could accelerate development of vision-language models and reduce focus on text-only model enhancements.

Cisco’s provenance tool release signals growing enterprise awareness of AI supply chain risks. As organizations deploy more third-party models, security and compliance tooling will become critical infrastructure requirements, potentially creating new market opportunities for AI governance solutions.

FAQ

What makes Poolside’s Laguna XS.2 different from other coding models?
Laguna XS.2 is optimized specifically for agentic workflows, meaning it can autonomously write code, use third-party tools, and take actions beyond simple code generation. Unlike many proprietary alternatives, it’s available as open source for local deployment.

Why are image AI models driving more app downloads than text models?
Image models provide more immediately visible and shareable results compared to text improvements. Users can create and share visual content more easily than demonstrating conversational AI capabilities, leading to viral adoption patterns that drive download growth.

How do US export controls affect Chinese AI model development?
US restrictions limit Chinese access to advanced AI training chips, forcing companies like SenseTime to optimize for domestic hardware. This has accelerated Chinese chip development and pushed firms toward more efficient model architectures that require less computational power.