Surprisingly Useful AI Article Enhancements
A new Australian AI company, Metacognition, has publicly launched at Deep Tech Week 2026 in Sydney, backed by Main Sequence Ventures and the South Australian Government, with plans to build AI systems that retain memory and reason over time rather than resetting with each interaction.
The company was founded by Professor Anton van den Hengel, Professor Stephen Gould and Dr Paul Dalby, who between them bring backgrounds spanning the Australian Institute for Machine Learning (AIML), Amazon and Stanford University.
Addressing The Limits Of Current AI
Metacognition’s pitch centres on what the founders describe as fundamental limitations in current large language models. The company argues that today’s LLMs are primarily text prediction systems that lack persistent memory and consistent reasoning, requiring users to effectively re-brief them with each new session.
The company is building a new architecture that includes memory that persists across interactions, reasoning that adapts over time and outputs that can be inspected for trustworthiness.
The practical implication, according to the founders, is AI that no longer needs to be retaught, can handle changing and unpredictable environments, and can be deployed in situations where the consequences of its outputs matter. The company is targeting enterprise systems and robotics as initial application areas.
Professor Stephen Gould, drew an analogy to workplace collaboration.
“Right now, using AI feels a bit like working with someone who’s briefed fresh every meeting. Even where AI does remember things, it’s usually a short list of facts without a real understanding of how multifaceted people are,” Gould explained.
“We think AI should feel like working with your favourite colleague, someone who gets better the longer you work together and adapts to and for you and the complex needs of human society. You shape how it thinks and works.”
Safety By Design
Professor Anton van den Hengel, emphasised that safety and human control are being built into the system’s architecture from the outset rather than added after the fact.
“We are prioritising building safety and human control into the design. That means people can stay in control at important decision points from day one, and it means safety is built into the system’s architecture,” van den Hengel noted. “We’re adopting Australia’s voluntary AI Safety Standard as our baseline and aiming to exceed it with our own internal benchmarks.”
Van den Hengel also framed the company as part of a broader ambition for Australian AI capability.
“Australia’s always punched above its weight in new technology and we think AI should be no different. We’re proud to be building that here.”
Backing And Talent
Main Sequence Ventures, a deep tech venture capital firm, is backing the company. Managing Partner, Bill Bartee, pointed to the limitations of current AI context windows as the problem Metacognition is attempting to solve.
“Right now, every model is limited by its context window – Metacognition is building the memory and reasoning layer that lets AI do genuinely hard, complex work over time,” Bartee commented. “Anton and Steve are generational experts who’ve built world-leading institutes in AIML and at ANU – together, they have exactly the calibre needed to solve this.”
Dalby, indicated the company is seeking further investment and looking to attract both Australian and international talent.
Related: Best Business Laptops for work & school
Related: Best Gaming Laptops
Related: Best Portable Laptop
“We want Australian talent at the centre of what we build, and we’re also hoping to attract the best global talent here to help us do it,” he remarked. “We’re grateful that Main Sequence and the South Australian Government backed us early, and we’re looking for more investors who believe in what we’re building.”
South Australian Government Support
South Australian Minister for Artificial Intelligence and Digital Economy, Chris Picton MP, described Metacognition as a product of the state government’s investment in AIML.
“It is precisely why the State Government made our foundational investment in the Australian Institute for Machine Learning – to build the next generation of nation-leading tech companies right here in SA,” Picton noted. “Our partnership with AIML is central to our ambition to build a globally competitive AI ecosystem, strengthen South Australia’s sovereign AI capability and create new economic opportunities for our State.”
Founder Backgrounds
Professor van den Hengel is Chief Scientist of AIML and its Founding Director, where he grew the institute to 130 researchers during a period when it was ranked second globally in computer vision. He was also appointed Director of Machine Learning at Amazon and created International Machine Learning Australia, which generates over US$1 billion per year. He has an H-index of 92, more than 400 papers and over 40,000 citations.
Professor Gould was previously Principal Research Scientist at Amazon and Director of the Centre of Excellence in Robotic Vision. He co-founded Sensory Networks, which was acquired by Intel in 2013, and received his PhD from Stanford University. He has an H-index of 62 and more than 26,000 citations.
Dr Dalby has spent 20 years working in research strategy and innovation, helping secure more than $600 million in research funding and establishing multiple research institutes and collaborative initiatives between industry, government and academia.
Last Updated on September 10, 2026 by Nick Ross
