b.well Launches the First-Ever SDK for Health AI Assistants
Organizations can now build AI assistants that understand a person’s full medical record and make it effortless to find care, manage medications, book appointments, and more b.well Connected Health, the most connected digital health platform for AI-powered consumer experiences, today unveiled the first SDK built to power healthcare AI assistants...
The post b.well Launches the First-Ever SDK for Health AI Assistants first appeared on AI-Tech Park.
Salesforce Buys Qualified in Agentic Marketing Push
The Government Added a Comments Section to the Epstein Photo Dump
Palona goes vertical, launching Vision, Workflow features: 4 key lessons for AI builders
Building an enterprise AI company on a "foundation of shifting sand" is the central challenge for founders today, according to the leadership at Palona AI.
Today, the Palo Alto-based startup—led by former Google and Meta engineering veterans—is making a decisive vertical push into the restaurant and hospitality space with today's launch of Palona Vision and Palona Workflow.
The new offerings transform the company’s multimodal agent suite into a real-time operating system for restaurant operations — spanning cameras, calls, conversations, and coordinated task execution.
The news marks a strategic pivot from the company’s debut in early 2025, when it first emerged with $10 million in seed funding to build emotionally intelligent sales agents for broad direct-to-consumer enterprises.
Now, by narrowing its focus to a "multimodal native" approach for restaurants, Palona is providing a blueprint for AI builders on how to move beyond "thin wrappers" to build deep systems that solve high-stakes physical world problems.
“You’re building a company on top of a foundation that is sand—not quicksand, but shifting sand,” said co-founder and CTO Tim Howes, referring to the instability of today’s LLM ecosystem. “So we built an orchestration layer that lets us swap models on performance, fluency, and cost.”
VentureBeat spoke with Howes and co-founder and CEO Maria Zhang in person recently at — where else? — a restaurant in NYC about the technical challenges and hard lessons learned from their launch, growth, and pivot.
The New Offering: Vision and Workflow as a ‘Digital GM’
For the end user—the restaurant owner or operator—Palona’s latest release is designed to function as an automated "best operations manager" that never sleeps.
Palona Vision uses in-store security cameras to analyze operational signals — such as queue lengths, table turnover, prep bottlenecks, and cleanliness — without requiring any new hardware.
It monitors front-of-house metrics like queue lengths, table turns, and cleanliness, while simultaneously identifying back-of-house issues like prep slowdowns or station setup errors.
Palona Workflow complements this by automating multi-step operational processes. This includes managing catering orders, opening and closing checklists, and food prep fulfillment. By correlating video signals from Vision with Point-of-Sale (POS) data and staffing levels, Workflow ensures consistent execution across multiple locations.
“Palona Vision is like giving every location a digital GM,” said Shaz Khan, founder of Tono Pizzeria + Cheesesteaks, in a press release provided to VentureBeat. “It flags issues before they escalate and saves me hours every week.”
Going Vertical: Lessons in Domain Expertise
Palona’s journey began with a star-studded roster. CEO Zhang previously served as VP of Engineering at Google and CTO of Tinder, while Co-founder Howes is the co-inventor of LDAP and a former Netscape CTO.
Despite this pedigree, the team’s first year was a lesson in the necessity of focus.
Initially, Palona served fashion and electronics brands, creating "wizard" and "surfer dude" personalities to handle sales. However, the team quickly realized that the restaurant industry presented a unique, trillion-dollar opportunity that was "surprisingly recession-proof" but "gobsmacked" by operational inefficiency.
"Advice to startup founders: don't go multi-industry," Zhang warned.
By verticalizing, Palona moved from being a "thin" chat layer to building a "multi-sensory information pipeline" that processes vision, voice, and text in tandem.
That clarity of focus opened access to proprietary training data (like prep playbooks and call transcripts) while avoiding generic data scraping.
1. Building on ‘Shifting Sand’
To accommodate the reality of enterprise AI deployments in 2025 — with new, improved models coming out on a nearly weekly basis — Palona developed a patent-pending orchestration layer.
Rather than being "bundled" with a single provider like OpenAI or Google, Palona’s architecture allows them to swap models on a dime based on performance and cost.
They use a mix of proprietary and open-source models, including Gemini for computer vision benchmarks and specific language models for Spanish or Chinese fluency.
For builders, the message is clear: Never let your product's core value be a single-vendor dependency.
2. From Words to ‘World Models’
The launch of Palona Vision represents a shift from understanding words to understanding the physical reality of a kitchen.
While many developers struggle to stitch separate APIs together, Palona’s new vision model transforms existing in-store cameras into operational assistants.
The system identifies "cause and effect" in real-time—recognizing if a pizza is undercooked by its "pale beige" color or alerting a manager if a display case is empty.
"In words, physics don't matter," Zhang explained. "But in reality, I drop the phone, it always goes down... we want to really figure out what's going on in this world of restaurants".
3. The ‘Muffin’ Solution: Custom Memory Architecture
One of the most significant technical hurdles Palona faced was memory management. In a restaurant context, memory is the difference between a frustrating interaction and a "magical" one where the agent remembers a diner’s "usual" order.
The team initially utilized an unspecified open-source tool, but found it produced errors 30% of the time. "I think advisory developers always turn off memory [on consumer AI products], because that will guarantee to mess everything up," Zhang cautioned.
To solve this, Palona built Muffin, a proprietary memory management system named as a nod to web "cookies". Unlike standard vector-based approaches that struggle with structured data, Muffin is architected to handle four distinct layers:
Structured Data: Stable facts like delivery addresses or allergy information.
Slow-changing Dimensions: Loyalty preferences and favorite items.
Transient and Seasonal Memories: Adapting to shifts like preferring cold drinks in July versus hot cocoa in winter.
Regional Context: Defaults like time zones or language preferences.
The lesson for builders: If the best available tool isn't good enough for your specific vertical, you must be willing to build your own.
4. Reliability through ‘GRACE’
In a kitchen, an AI error isn't just a typo; it’s a wasted order or a safety risk. A recent incident at Stefanina’s Pizzeria in Missouri, where an AI hallucinated fake deals during a dinner rush, highlights how quickly brand trust can evaporate when safeguards are absent.
To prevent such chaos, Palona’s engineers follow its internal GRACE framework:
Guardrails: Hard limits on agent behavior to prevent unapproved promotions.
Red Teaming: Proactive attempts to "break" the AI and identify potential hallucination triggers.
App Sec: Lock down APIs and third-party integrations with TLS, tokenization, and attack prevention systems.
Compliance: Grounding every response in verified, vetted menu data to ensure accuracy.
Escalation: Routing complex interactions to a human manager before a guest receives misinformation.
This reliability is verified through massive simulation. "We simulated a million ways to order pizza," Zhang said, using one AI to act as a customer and another to take the order, measuring accuracy to eliminate hallucinations.
The Bottom Line
With the launch of Vision and Workflow, Palona is betting that the future of enterprise AI isn't in broad assistants, but in specialized "operating systems" that can see, hear, and think within a specific domain.
In contrast to general-purpose AI agents, Palona’s system is designed to execute restaurant workflows, not just respond to queries — it's capable of remembering customers, hearing them order their "usual," and monitoring the restaurant operations to ensure they deliver that customer the food according to their internal processes and guidelines, flagging whenever something goes wrong or crucially, is about to go wrong.
For Zhang, the goal is to let human operators focus on their craft: "If you've got that delicious food nailed... we’ll tell you what to do."
Anthropic Launches Skills Open Standard for Claude
ICE Contracts Company Making Bounty Hunter AI Agents
Google Releases Gemini 3 Flash Aimed at Enterprises
Anthropic launches enterprise ‘Agent Skills’ and opens the standard, challenging OpenAI in workplace AI
Anthropic said on Wednesday it would release its Agent Skills technology as an open standard, a strategic bet that sharing its approach to making AI assistants more capable will cement the company's position in the fast-evolving enterprise software market.
The San Francisco-based artificial intelligence company also unveiled organization-wide management tools for enterprise customers and a directory of partner-built skills from companies including Atlassian, Figma, Canva, Stripe, Notion, and Zapier.
The moves mark a significant expansion of a technology Anthropic first introduced in October, transforming what began as a niche developer feature into infrastructure that now appears poised to become an industry standard.
"We're launching Agent Skills as an independent open standard with a specification and reference SDK available at https://agentskills.io," Mahesh Murag, a product manager at Anthropic, said in an interview with VentureBeat. "Microsoft has already adopted Agent Skills within VS Code and GitHub; so have popular coding agents like Cursor, Goose, Amp, OpenCode, and more. We're in active conversations with others across the ecosystem."
Inside the technology that teaches AI assistants to do specialized work
Skills are, at their core, folders containing instructions, scripts, and resources that tell AI systems how to perform specific tasks consistently. Rather than requiring users to craft elaborate prompts each time they want an AI assistant to complete a specialized task, skills package that procedural knowledge into reusable modules.
The concept addresses a fundamental limitation of large language models: while they possess broad general knowledge, they often lack the specific procedural expertise needed for specialized professional work. A skill for creating PowerPoint presentations, for instance, might include preferred formatting conventions, slide structure guidelines, and quality standards — information the AI loads only when working on presentations.
Anthropic designed the system around what it calls "progressive disclosure." Each skill takes only a few dozen tokens when summarized in the AI's context window, with full details loading only when the task requires them. This architectural choice allows organizations to deploy extensive skill libraries without overwhelming the AI's working memory.
Fortune 500 companies are already using skills in legal, finance, and accounting
The new enterprise management features allow administrators on Anthropic's Team and Enterprise plans to provision skills centrally, controlling which workflows are available across their organizations while letting individual employees customize their experience.
"Enterprise customers are using skills in production across both coding workflows and business functions like legal, finance, accounting, and data science," Murag said. "The feedback has been positive because skills let them personalize Claude to how they actually work and get to high-quality output faster."
The community response has exceeded expectations, according to Murag: "Our skills repository already crossed 20k stars on GitHub, with tens of thousands of community-created and shared skills."
Atlassian, Figma, Stripe, and Zapier join Anthropic's skills directory at launch
Anthropic is launching with skills from ten partners, a roster that reads like a who's who of modern enterprise software. The presence of Atlassian, which makes Jira and Confluence, alongside design tools Figma and Canva, payment infrastructure company Stripe, and automation platform Zapier suggests Anthropic is positioning Skills as connective tissue between Claude and the applications businesses already use.
The business arrangements with these partners focus on ecosystem development rather than immediate revenue generation.
"Partners who build skills for the directory do so to enhance how Claude works with their platforms. It's a mutually beneficial ecosystem relationship similar to MCP connector partnerships," Murag explained. "There are no revenue-sharing arrangements at this time."
For vetting new partners, Anthropic is taking a measured approach. "We began with established partners and are developing more formal criteria as we expand," Murag said. "We want to create a valuable supply of skills for enterprises while helping partner products shine."
Notably, Anthropic is not charging extra for the capability. "Skills work across all Claude surfaces: Claude.ai, Claude Code, the Claude Agent SDK, and the API. They're included in Max, Pro, Team, and Enterprise plans at no additional cost. API usage follows standard API pricing," Murag said.
Why Anthropic is giving away its competitive advantage to OpenAI and Google
The decision to release Skills as an open standard is a calculated strategic choice. By making skills portable across AI platforms, Anthropic is betting that ecosystem growth will benefit the company more than proprietary lock-in would.
The strategy appears to be working. OpenAI has quietly adopted structurally identical architecture in both ChatGPT and its Codex CLI tool. Developer Elias Judin discovered the implementation earlier this month, finding directories containing skill files that mirror Anthropic's specification—the same file naming conventions, the same metadata format, the same directory organization.
This convergence suggests the industry has found a common answer to a vexing question: how do you make AI assistants consistently good at specialized work without expensive model fine-tuning?
The timing aligns with broader standardization efforts in the AI industry. Anthropic donated its Model Context Protocol to the Linux Foundation on December 9, and both Anthropic and OpenAI co-founded the Agentic AI Foundation alongside Block. Google, Microsoft, and Amazon Web Services joined as members. The foundation will steward multiple open specifications, and Skills fit naturally into this standardization push.
"We've also seen how complementary skills and MCP servers are," Murag noted. "MCP provides secure connectivity to external software and data, while skills provide the procedural knowledge for using those tools effectively. Partners who've invested in strong MCP integrations were a natural starting point."
The AI industry abandons specialized agents in favor of one assistant that learns everything
The Skills approach is a philosophical shift in how the AI industry thinks about making AI assistants more capable. The traditional approach involved building specialized agents for different use cases — a customer service agent, a coding agent, a research agent. Skills suggest a different model: one general-purpose agent equipped with a library of specialized capabilities.
"We used to think agents in different domains will look very different," Barry Zhang, an Anthropic researcher, said at an industry conference last month, according to a Business Insider report. "The agent underneath is actually more universal than we thought."
This insight has significant implications for enterprise software development. Rather than building and maintaining multiple specialized AI systems, organizations can invest in creating and curating skills that encode their institutional knowledge and best practices.
Anthropic's own internal research supports this approach. A study the company published in early December found that its engineers used Claude in 60% of their work, achieving a 50% self-reported productivity boost—a two to threefold increase from the prior year. Notably, 27% of Claude-assisted work consisted of tasks that would not have been done otherwise, including building internal tools, creating documentation, and addressing what employees called "papercuts" — small quality-of-life improvements that had been perpetually deprioritized.
Security risks and skill atrophy emerge as concerns for enterprise AI deployments
The Skills framework is not without potential complications. As AI systems become more capable through skills, questions arise about maintaining human expertise. Anthropic's internal research found that while skills enabled engineers to work across more domains—backend developers building user interfaces, researchers creating data visualizations—some employees worried about skill atrophy.
"When producing output is so easy and fast, it gets harder and harder to actually take the time to learn something," one Anthropic engineer said in the company's internal survey.
There are also security considerations. Skills provide Claude with new capabilities through instructions and code, which means malicious skills could theoretically introduce vulnerabilities. Anthropic recommends installing skills only from trusted sources and thoroughly auditing those from less-trusted origins.
The open standard approach introduces governance questions as well. While Anthropic has published the specification and launched a reference SDK, the long-term stewardship of the standard remains undefined. Whether it will fall under the Agentic AI Foundation or require its own governance structure is an open question.
Anthropic's real product may not be Claude—it may be the infrastructure everyone else builds on
The trajectory of Skills reveals something important about Anthropic's ambitions. Two months ago, the company introduced a feature that looked like a developer tool. Today, that feature has become a specification that Microsoft builds into VS Code, that OpenAI replicates in ChatGPT, and that enterprise software giants race to support.
The pattern echoes strategies that have reshaped the technology industry before. Companies from Red Hat to Google have discovered that open standards can be more valuable than proprietary technology — that the company defining how an industry works often captures more value than the company trying to own it outright.
For enterprise technology leaders evaluating AI investments, the message is straightforward: skills are becoming infrastructure. The expertise organizations encode into skills today will determine how effectively their AI assistants perform tomorrow, regardless of which model powers them.
The competitive battles between Anthropic, OpenAI, and Google will continue. But on the question of how to make AI assistants reliably good at specialized work, the industry has quietly converged on an answer — and it came from the company that gave it away.