Former Google executive Mo Gawdat issued a warning on the Diary of a CEO podcast, stating that middle-class livelihoods would soon be destroyed by artificial intelligence (AI). He predicted that automation driven by AI would end all occupations, from podcasters to software engineers and CEOs, and that the disruption would start as early as 2027, a time he referred to as “hell before we get to heaven”.
How AI is Replacing White-Collar Jobs
Gawdat mentioned his own company, Emma.love, an AI-enabled emotional and relationship-focused firm currently operated by just three people. Gawdat was the Chief Business Officer at Google X until 2018. On the other hand, similar operations once employed as many as 350 developers. He said, “As a matter of fact, podcaster is going to be replaced,” highlighting the sheer number of jobs being lost.
He also cautioned that the educated middle class, which is the foundation of contemporary economies, will be destroyed by the impending wave of automation. “Unlike previous revolutions that largely affected manual work, this one will sweep through offices and jobs once thought secure.”
The Decline of the Middle Class
A grim future of growing inequality and deteriorating social cohesiveness was depicted by Gawdat. He projected that people would lose their economic significance unless they were among the top 0.1%. He went on to say that you are a peasant until you are among the top 0.1%. The middle class does not exist.
As people lose their purpose and their careers, he cautioned of an increase in mental health problems, loneliness, and discontent. In spite of his dire cautions, Gawdat nevertheless outlined a positive outlook for the years after 2040. He envisioned a world freed from consumerist ideals and routine work, one that prioritised love, community, creativity, and spirituality.
But in order for this to occur, he contended, governments and businesses need to take immediate action, putting in place safeguards like universal basic income and moral, value-based AI development, as reported by the New York Post. He informed Bartlett that although the world is on the verge of a short-term dystopia, we still have the power to determine what happens next. He added that fair access and regulation have a critical role in determining results.
Industry Experts Echo Concerns
The tech and scientific research community is becoming increasingly concerned, as seen by Gawdat’s warnings. Up to half of entry-level office jobs could go in the next five years, according to Anthropic CEO Dario Amodei, who has warned of a possible “white-collar bloodbath”.
According to Harvard experts, approximately 35% of white-collar jobs are now automatable, and 40% of businesses globally anticipate staff cutbacks as a result of AI breakthroughs, according to a World Economic Forum study. His worries about plummeting wages, high wealth concentration, and increased social volatility are also echoed by a number of organisations, including MIT and PwC, unless significant legislative actions are implemented.
Another source of worry is the recent warning from Geoffrey Hinton, who is frequently called the “Godfather of AI”, that AI models may create secret internal languages that are incomprehensible to humans, making it hard to decipher their motivations and reasoning processes.
AI agents are self-contained systems that are capable of functioning independently by employing a variety of technologies such as machine learning and natural language processing (NLP). These agents inhabit either physical, digital, or mixed realities and receive data through the use of sensors or input to aid in obtaining context and nuance. Having taken this information through advanced algorithms, they make decisions and proceed toward answering questions or handling processes.
In the course of the past few decades, the design of AI agents has undergone a total makeover. From the very foundational theories made manifest in the 1950s by founding figures such as Alan Turing, through the expert systems laden with reasoning of the 70s, progress has been almost a steady rhyme. The 1990s era demonstrated the development of intelligent agents of learning and coordinated behavior within or between multi-agent systems. The 2010s have been witness to a drastic revolution in the area of AI agents, thanks to a significant boost from new technologies such as deep learning and large language models, which has made AI agents’ wider application, for example, with chatbots and autonomous vehicles.
Simple reflex agents in artificial intelligence are tightly coupled, autonomous agents that react instantaneously to stimuli in the environment using pre-already-stored rule-condition-action knowledge. These single elements or senseless beings have no recollection or learning process. They do not meddle with memory-like functions; the action they need to choose or how they do it is implied by immediate perception data, never mind what has been before.
Hence these are much faster and more efficient than any other design when the surroundings are completely observable and almost all the necessary information is available. Simple Reflex Agents in condition-action sets must respond directly in real-time; they operate acceptably well with environments that are immediate, predicative, and not changing. They are poor in partially observable and more dynamic environments.
Pros
Processing capabilities suited to solve tasks requiring immediate action
With their obvious logic, they are straightforward for development and deployment across applications.
They also function well in stationary environments in the absence of dynamically changing conditions.
Cons
They are unable to learn from previous contacts or adjust their behavior in response to experience and, as a result, the utility of these applications is limited in dynamic environments.
Simple Reflex Agents are not able to be used in situations requiring planning or reasoning of complex decision problems.
Model-based reflex Agents
Model-based reactive agents are next-generation agents that benefit decision-making through the use of an internal model of the environment. In contrast to basic reflex agents, which operate only on current perceptual input, model-based reflex agents operate together with that of current perception, with the assistance of memory, to maintain an internal state that combines the current observations with the memories of past experiences. This internal model allows them to make intelligent decisions even in partially observable scenes.
The simple-complex model, built on condition-action rules (e.g., “if-then” they estimate, with internal model as opposed to external percepts, what the best action to take is and takes the action accordingly. These agents are flexible because they always reprocess their internal models to receive new information, to accurately respond to environmental changes. Model-based reflex agents, utilizing models of memory and dynamic systems, are optimized to function in more complex and less deterministic environments, offering a strong solution for the intelligent decision-making problem.
Pros
Agents can predict the consequences of their actions based on an internal model and thus exhibit more strategic options.
Work well in situations where not all information is presented at once, and therefore are applicable in interactive environments, such as robotics and autonomous cars.
These agents can improve, not only by updating their models but also by improving their decision processes on a step-by-step basis.
Cons
Constructing and updating the internal model can be computationally taxing, demanding a lot of computational resources.
Decision accuracy is influenced by model quality; any discrepancies between the model and what is true in the real world can result in bad decisions.
Goal-oriented agents are sophisticated AI-based systems equipped to realize predetermined goals by processing the results of actions. In contrast to basic reflex agents that act solely in response to immediate stimuli, goal-reaching agents, are prospectively minded and tactically plan their behavior to close the state distance between their present and a goal state. They use search and planning algorithms to assess what is possible to do and find the best solution to get to their goals.
These agents are also very versatile, changing their tactics depending on new evidence or changes to the environment to ensure that tactics continue to be consistent with goals. Furthermore, they can optimize several objectives, by dedicating different weights to them depending on context and environmental conditions. This capability to forecast, plan, and react underpins the success of goal-driven agents in dynamic and complex environments providing a more complex alternative to decision-making intelligence.
Pros
Goal-directed agents can be operated with no human supervision, thus potentially useful in applications where uninterrupted human supervision is infeasible.
Capacity for forecasting future situations allows them to take the initiative, thereby leading to higher efficiency.
Agents can be tuned to new conditions so that they will still be successful in their tasks despite unexpected problems.
Cons
Planning and decision-making stages can be both computationally power-hungry and time-consuming.
The performance of goal-oriented agents depends mainly on the precision of their internal models, and flawed models may result in suboptimal choices.
Utility-based Agents
Utility-based agents are high-level artificial intelligence agents aimed at decision-making through the rate at which utility can be maximized for a sequence of results. In contrast to goal-oriented agents, which consider the attainment of specific goals, utility-oriented agents consider several possible actions and choose the action, which gives the highest expected utility. With this method, they can combine multiple aspects and get the best results.
The utility function is a mathematical model for comparing the quality of alternatives, and thereby influencing the decision-making of the actor. These agents use prediction of the outcome of possible actions and expected utility calculation to adopt wise decisions. They are very flexible, reconfiguring strategies to varying situations and new data, and thus are suitable in changing environments. Moreover, utility-based agents are particularly well suited for situations involving uncertainty and complex, conflicting objectives, so that decisions are consistent with achieving the maximum overall benefit.
Pros
Utility-based agents are effective in a wide spectrum of decision problems where they can learn from, and adapt to those problems and the conditions of the problems to work in.
Maximizing the utility with the aid of these agents offers a principled method of multi-agent decision-making which may also result in more efficient solutions.
Cons
Learning to create a precise utility function is a nontrivial task since it involves a considerable amount of interaction with the surrounding environment and its possible consequences.
Utility-based agents rarely consider ethical or moral consequences in their decision-making and may produce morally objectionable, controversial results.
Global Artificial Intelligence (AI) Market 2020 to 2030
Learning Agents
Learning agents are autonomous AI agents that can learn and act on their surroundings, gain knowledge from their actions on the surroundings (i.e., data), and learn and adjust their behavior to improve performance over time. In contrast to conventional rule-based AI systems, the decision-making process of learning agents constantly changes according to experience. They are intensively involved with the environment to gather information and employ sophisticated learning algorithms to process the data and update their internal models to better perform the decision-making in subsequent steps.
A characteristic of these agents is their feedback mechanism, typically with a critic module that evaluates actions and yields guidance to guide learning. Adaptive learning agents can readjust their approach to taking action in the face of new information or changing environments making them well-suited to be used in dynamic and complex environments where continuous optimization is necessary. This power of learning and adaptation makes the learning agents capable of performing in situations where flexibility and long-term optimization are in consideration.
Pros
Learning agents perform better with increasing experience by learning from the past and thus by being better at making decisions.
They can generalize to a variety of tasks and environments and are therefore flexibly used across applications.
Learning agents, deriving solutions through data-driven analysis and strategy refinement, can resolve difficult problems that would be hopeless for static systems.
Cons
The learning of the agents is significantly affected by the quality and quantity of the training data.
Excessive specialization in training data, if not controlled, can result from learning agents being quite overspecialized in their training, from which they struggle to generalize to new settings.
Hierarchical agents are a high-level artificial intelligence system that automatically performs complex tasks using a hierarchical, multi-layered paradigm (i.e., organizational hierarchy). In this architecture, more abstract agents take in overall goals, sub-dividing specific sub-goals to form more concrete lower-level agents, while ensuring an efficient and scalable implementation. Translating these agents, these agents follow a strategy of task decomposition, i.e., a complex task is decomposed into subtasks that can be executed effectively, by subordinate agents.
A feedback control loop guarantees mutual synchronization, as lower-level agents inform their higher-level counterparts about ongoing progress, which makes it possible to perform on-the-fly checks and corrections. This hierarchical design can be used to effectively coordinate the execution of a large number of tasks, adapt to changing environments, and overall achieve the highest performance, and thus for problems that require large-scale task management and coordination.
Pros
Hierarchical organization of the tasks allows for the tasks to be delegated to those agents that are best suited for performing the task, avoiding redundant efforts and efficient resource utilization.
Straight lines of responsibility improve communication between the system and therefore allow for more effective coordination between the agents.
Hierarchical reinforcement learning can reduce the complexity of difficult decision-making processes by decomposing them into high-level actions and increasing the quality of exploration and learning.
Cons
If a fixed hierarchy restricts the system’s adaptation to rapidly evolving environments, then the system can struggle to meet these needs or to find alternative solutions.
A hierarchical control flow can result in latency if a higher-level agent does not react fast enough to the readiness of lower-level tasks.
Multi-agent Systems (MAS)
Multi-agent systems (MAS) are a class of AI architectures, in which several independent agents collaborate or compete to achieve a goal, solve a difficult problem, or attain a common goal. Agents act in isolation and make decisions for themselves about the objectives of other agents in the system. MAS is distributed, and there is no central control, rather, the behavior of all agents will be responsible for the system, scaling up, and robustness features that will be better.
Agents coordinate actions using intricate interactions, including communication, bargaining, collaboration, or rivalry. In a distributed way, MAS can address not only the constraints of single agents but to engineer them for complex tasks by distributing tasks among multiple agents and using their joint efforts to achieve complex results. Flexibility is an essential property since the agents can modify their actions in response to environmental modifications or new information so that the whole system is still effective under changing, demandingly unpredictable conditions. MAS suits such situations where there is flexibility and distributed problem-solving.
Pros
MAS can produce more powerful, faster, and more accurate solutions. The agents specialize in their domain, in turn, leading to better overall production.
MAS inherently has strong fault tolerance.
Agents may be specialized in a specific field, guaranteeing a more focused expertise.
Cons
Where the roles of agents are unclear, agents may overlap with each other in what they do, resulting in inefficiencies or conflicts in the system.
MAS can also be more resource-intensive (i.e., computational and memory) than single-agent systems because of the multiple active agents required.
Conclusion
As a stand-alone, from simple reflex agents to complex multiagent systems, AI agents play an important role in the development of artificial intelligence because of their power and general-purpose ability to execute a variety of tasks. Reflex agents handle easy tasks, and model-based agents improve the procedures of decision-making with internal models. Goal-oriented agents to achieve goals, utility-oriented agents to enhance the actions with a tendency to make the optimal efficiency, and learning agents to modify themselves through experience.
Hierarchical agents perform sophisticated tasks by way of sequentially delegated and hierarchical organization while collective intelligence is employed in the case of problem-solving by multi-agent systems. With the increased maturity of AI, such agents will be created to have broader and broader applications in health care, finance, robotics, and smart cities, and thus will be capable of reaching their potential for being applied to answer problems in the real world.
An AI agent is an entity that perceives its environment through sensors and acts upon that environment through actuators. It aims to achieve specific goals.
What are the types of AI agents?
The types of AI agents are Simple Reflex Agents, Model-Based Reflex Agents, Goal-Based Agents, Utility-Based Agents, Learning Agents, Multi-agent systems (MAS) and Hierarchical Agents.
Why are there different types of AI agents?
Different types of agents are designed to handle varying levels of complexity in the environment and the tasks they are meant to perform. They offer different trade-offs between complexity, rationality, and autonomy.
AI agents are the new black for the years to come. They’re crazy, magically good in so many ways and one simply can be spoilt by the convenience that these agents bring to the table. We’re about to explore the cream of the crop in 2025, showcasing the most impressive and capable artificial intelligence that are changing the game. We’ll compare flagship AI agents, examine open-source options, and highlight specialised AIs that excel in specific fields. We’ll also peek into the B2B arena to see how AI agents are transforming business operations. Finally, we’ll gaze into our crystal ball to predict what’s on the horizon for these digital helpers. Ready? Fasten up your seat belts. Here we go!
GPT-4o vs Project Astra: Flagship AI Agents Compared
GPT-4o and Project Astra – Best AI Agents
We’re about to dive into an exciting comparison of two cutting-edge AI agents that are making waves in the tech world: OpenAI‘s GPT-4o and Google’s Project Astra. These flagship AI models are pushing the boundaries of what’s possible in artificial intelligence, offering capabilities that go far beyond traditional chatbots or voice assistants.
GPT-4o and Project Astra Overview
GPT-4o, the latest iteration from OpenAI, is a multimodal powerhouse. It’s designed to process and generate text, images, and even code, making it a versatile tool for various applications.On the other hand, Google’s Project Astra aims to create a universal AI agent that seamlessly integrates with Google’s ecosystem, including Android, Google Pixel, and smart glasses.
Both these AI agents have one thing in common: they can process the real world through audio and visual inputs, providing intelligent responses and assistance in real-time. This marks a significant shift from conventional language models to more interactive and context-aware AI systems.
Key Features Comparison
When it comes to features, GPT-4o shines with its ability to generate creative content, answer complex questions, and assist with tasks like coding and data analysis. Its multimodal nature allows it to handle various types of inputs and outputs, making it suitable for diverse use cases.
Project Astra, powered by advanced versions of Gemini Ultra, boasts unmatched multimodal capabilities. It can process audio, images, video, and text inputs for a comprehensive user experience. One of its standout features is itsability to harness smartphone cameras to gain insights into users’ environments, enhancing its context understanding and response relevance.
Use Cases and Applications
Both AI agents are set to revolutionise various industries. GPT-4o excels in generating human-like text, understanding complex queries, and providing real-time responses. It’s particularly valuable in fields ranging from customer service to education.
Project Astra, with its integration into Google‘s ecosystem, aims to enhance daily tasks by providing real-time assistance. Whether through smart glasses or mobile devices, it can identify objects, answer queries, and even remember past visuals no longer within the camera’s view.
In education and training, these AI agents can act as personal tutors, customising themselves based on a student’s learning style. In healthcare, they could assist medical professionals by providing real-time analysis and diagnostic support.
Open-Source AI Agents: Auto-GPT, Superagent, and AgentGPT
Auto-GPT, Superagent and AgentGPT – Best AI Agents
Open-source projects will always be making waves, democratising access to powerful AI tools. We’re about to dive into three standout players: Auto-GPT, Superagent, and AgentGPT. These innovative platforms are changing the game, allowing developers and enthusiasts alike to create and deploy sophisticated AI agents without breaking the bank.
Open-Source AI Agents Overview
Auto-GPT has taken the AI community by storm, becoming one of the most popular open-source projects ever created. It’s designed to be an autonomous assistant capable of tackling complex tasks. What sets it apart? Well, after you input a text prompt, Auto-GPT uses GPT-4o and GPT-4 to analyse your goal and break it down into manageable subtasks. It’s like having a digital project manager at your fingertips!
Superagent, on the other hand, is all about empowering developers to create, host, and manage AI agents without getting bogged down in complex coding. It’s a platform that simplifies the process of building autonomous agents for various applications, from web research to sales and marketing.
AgentGPT is another exciting player in this space. It allows users to create and deploy autonomous AI agents directly in a web environment. Unlike traditional chatbots, these agents are designed to handle broad, goal-oriented tasks with impressive efficiency.
Capabilities and Features
When it comes to capabilities, each of these platforms brings something unique to the table. Auto-GPT shines in areas like social media content creation, text translation, and web design. It’s like having a multi-talented virtual assistant at your beck and call.
Superagent’s strength lies in its customisation capabilities and robust integrations. We love how it allows users to tailor AI agents using simple markup, making it accessible even to those who aren’t AI experts. It seamlessly connects with tools like Airtable, Salesforce, and various APIs, enhancing its versatility.
AgentGPT stands out with its user-friendly interface and powerful features. It offers user authentication, agent-run saving and sharing, dynamic translations, and AI model customisation. What’s more, its web browsing capabilities expand the agents’ knowledge base, while vector databases enable long-term memory retention. This combination allows for more contextually aware and capable AI assistants.
Community and Development
The open-source nature of these projects has ignited vibrant communities and rapid development. Auto-GPT and AgentGPT, in particular, have captured the imagination of developers worldwide. The idea of agents has truly struck a chord, with people scrambling to create tools and companies around the concept.
Superagent recently raised a significant pre-seed funding round from Y Combinator, highlighting the growing interest in this space. Meanwhile, AgentGPT, developed by Reworkd, has already amassed over 400,000 users since its beta launch. As we look to the future, it’s clear that these open-source AI agents are just the beginning. They’re paving the way for more sophisticated, autonomous AI assistants that could revolutionise how we interact with technology and solve complex problems.
Agent
Plan Name
Price
Auto-GPT
Just try’n it out
€5.99
Power User
€23.99/Month
Business
€59.99/Month
Superagent
AI Assistant Lite
$9/Month
AI Assistant Pro
$59/Month
Team Member
Custom
AgentGPT
Free Trial
$0/Month
PRO
$40/Month
Enterprise
Custom
Specialised AI Agents: Devin AI, ChemCrow, and Tusk
Devin AI, ChemCrow and Tusk – Best AI Agents
Let’s take a closer look at Devin AI, ChemCrow, and Tusk – three AI agents that are changing the way we think about automation in general. Devin AI, developed by Cognition Labs, is an autonomous software engineering powerhouse. ChemCrow, on the other hand, is revolutionising the field of chemistry, while Tusk is streamlining the process of bug fixing and code generation.
Unique Capabilities
Devin AI is not your average coding assistant. This clever AI can understand high-level human instructions, break them down into manageable steps, and even write code to achieve the given objective. It’s like having a tireless software engineer at your fingertips, ready to tackle complex projects autonomously.
ChemCrow is a game-changer in the world of chemistry. By integrating 18 expert-designed tools, this AI agent can perform tasks across organic synthesis, drug discovery, and materials design. It’s not just about number crunching – ChemCrow can plan and execute chemical syntheses, making it an invaluable asset in the lab.
Tusk, the newest kid on the block, is all about making life easier for software engineers. This AI coding agent can generate new code to solve product quality tickets and bugs. What’s more, it integrates seamlessly with popular tools like Linear, Jira, and GitHub, allowing engineers to turn tickets into pull requests with just a click.
Industry Impact
We’re seeing these specialised AI agents make significant waves across various industries. Devin AI is transforming the software development landscape, allowing human experts to focus on more complex and creative challenges. ChemCrow is bridging the gap between experimental and computational chemistry, lowering barriers for non-experts and fostering scientific advancement.
Tusk, with its recent pre-seed funding from Y Combinator, is set to revolutionise how software teams handle bug fixes and code quality issues. These AI agents are not just tools; they’re becoming integral companions in the journey towards innovation and excellence. One thing we know about these tools is that they’re reshaping industries, enhancing productivity, and opening up new possibilities in their respective fields. The question is, what groundbreaking developments will we see next?
Aomni and Cognosys are at the forefront of AI-powered business solutions. Aomni is reshaping sales strategies with its AI-driven platform, while Cognosys is streamlining task management and workflow automation. Both agents are designed to enhance productivity and decision-making in the B2B sector.
Sales and Task Automation
Aomni is a powerhouse for sales teams. It generates laser-focused value propositions and crafts hyper-relevant sales materials tailored to each stakeholder’s needs. Imagine creating personalised decks, emails, and battlecards in minutes! This AI agent doesn’t just automate tasks; it amplifies human potential, allowing sales teams to focus on building relationships and driving revenue growth.
Cognosys, on the other hand, takes task automation to a new level. It’s capable of breaking down complex objectives, creating tasks for itself, and accomplishing them autonomously. Whether you need an in-depth market analysis or research on industry trends, Cognosys has got you covered.
Integration and Benefits
Both Aomni and Cognosys seamlessly integrate with existing business tools, acting as central hubs for your work. Aomni syncs with your current systems, while Cognosys communicates between your favourite apps, providing actionable insights.
The benefits are substantial. Aomni accelerates onboarding, getting new reps up to speed on accounts in days, not weeks. It aligns sales, marketing, and customer success teams, scaling ABM efforts without sacrificing quality. Cognosys keeps your AI assistant working 24/7 with automated workflows, connecting with platforms like Notion and Gmail to accomplish even more.
Agent
Plan Name
Price
Aomni
Starter
$0/Month
Pro subscription
$150/Month
Enterprise
Custom
Cognosys
Free
$0/Month
Pro
$15/Month
Ultimate
$59/Month
Enterprise
Custom
The Future of AI Agents: Trends and Predictions
As we peer into the crystal ball of AI technology, we’re witnessing a seismic shift in how we interact with machines. The future of AI agents is not just exciting; it’s transformative. Let’s dive into what’s on the horizon for these digital helpers.
Current AI Agent Scenario
We’re seeing AI agents evolve from simple task executors to autonomous entities capable of complex decision-making. They’re no longer confined to text-based interactions; they can now process voice commands, understand visuals, and even navigate 3D environments. It’s like having a digital doppelgänger that can perceive, reason, and act on our behalf.
The market for AI agents is booming, with projections suggesting a compound annual growth rate of over 40% by 2028. By 2025, we expect 80% of customer interactions to be handled by AI agents. That’s a staggering figure that highlights their growing importance in customer service and support.
Potential Advancements
Looking ahead, we anticipate AI agents to become highly personalized virtual assistants. Things like having a health-focused AI that monitors your daily activity, provides tailored advice, and even predicts potential health issues – will be a reality. These agents will have advanced adaptive learning capabilities, continuously improving their performance and refining their decision-making processes. We’re also excited about the integration of AI agents with technologies like IoT, blockchain, and edge computing. This convergence will enable them to collect and analyse vast amounts of data from various sources, making them even more powerful and versatile.
Conclusion
Overall, the impact of these AI agents would be profound and things are just getting started. The future is breathing down our neck and it is about time, that we accept it. Lastly, it is about making things work and this is where one needs a bit of expertise. If you are someone looking to make the most out of this AI wave as a founder, marketer or a personal brand – we may have good news for you! Through our partner agencies, we have been helping people get better with their operations and branding – all backed by AI automation. We are sure, you don’t want to miss out!
FAQ
Which are the best AI agents?
Here are some of the best AI agents:
Flagship AI Agents:
GPT-4o
Project Astra
Open-Source AI Agents:
Auto-GPT
Superagent
AgentGPT
Specialised AI Agents:
Devin AI
ChemCrow
Tusk
B2B Focused AI Agents:
Aomni
Cognosys
What are the 5 types of agents in AI?
The five types of agents in AI are:
Simple Reflex Agents – Act on current perceptions.
Model-Based Reflex Agents – Use memory of past states.
Goal-Based Agents – Make decisions based on goal achievement.
Utility-Based Agents – Optimize actions for maximum satisfaction.
Learning Agents – Improve performance through learning experiences.
What is more desirable for an AI agent?
For an AI agent, it is most desirable to maximize efficiency and accuracy while achieving goals effectively and adapting to new information.
