Top news
Investing in the public benefit OfficialAnthropic NewsOct 18Popular Musk Pledges SpaceX Shares to ‘Every Kid in America’ WebBloomberg TechnologyOct 19Popular Websites are closing their doors to AI agents WebThe Verge AIOct 17 This is the new ChatGPT Voice, powered by GPT Live OfficialOpenAI on YouTubeOct 18 How AI giants will reshape global philanthropy WebAxios AIOct 19 Gemini 4 Argon Really Surprised Me WebTwo Minute PapersOct 19 Google Gemini 4 agents just beat Minecraft with no human intervention WebEngadget AIOct 19 Who could it be? 🧐 OfficialGoogle on YouTubeOct 18 Anthropic gives outside AI watchdog “employee-like” access WebArs Technica AIOct 18 What are self-replicating ‘AI worms’, and why is Anthropic worried about them? WebThe Guardian AIOct 18
GlossaryAgentsIntermediate

Long-horizon tasks

Also written: Long-horizon workflows, Long-horizon reasoning, Long context reasoning, Task Horizons, Long-time-horizon threshold, Long horizon planning, Long-horizon coding, Long-horizon task planning

Tasks that require extended planning, execution, and memory over a significant duration or many steps.

In plain terms

Long-horizon tasks are jobs that take a lot of time or involve many steps to complete. Think of building a complex piece of software or conducting a scientific experiment that requires days of work. These tasks need AI systems to remember what they've done, plan ahead, and keep working towards a goal over a long period.

For example, an AI agent might be asked to design a website. This involves many steps: understanding the user's needs, generating code, testing it, making revisions, and deploying it. The AI needs to manage all these steps over hours or even days to finish the job.

The expert version

Long-horizon tasks are complex computational problems that demand sustained planning, execution, and memory retention over extended periods or through a large number of sequential steps. These tasks often involve iterative refinement, interaction with external environments, and the management of intermediate states. Successfully completing them requires AI systems capable of maintaining context, adapting to new information, and coordinating multi-step processes autonomously.

Why it matters now

Recent advancements in large context windows, like those in Google's Gemini 4 Argon, are enabling AI models to process and retain information over much longer durations. This capability is crucial for tackling complex, multi-step tasks that previously exceeded AI's practical limits, leading to new frameworks for improving reliability in these extended operations.

Where it shows up 22 items, newest first
How each company uses it
Anthropic Claude
Anthropic's VeriHarness framework specifically addresses scaling agentic verification for long-horizon tasks, improving reliability.
Google Gemini
Google's Gemini 4 Argon model and Gemini 3.5 Flash are highlighted for their massive context windows and efficiency in handling complex, multi-step AI tasks and workflows.
OpenAI ChatGPT
OpenAI's GPT-6 Astra is mentioned in the context of enabling developers to ship features, implying assistance with longer development cycles or complex coding tasks.
Related terms

AI AgentsAutonomous SystemsTask DecompositionPlanningContext WindowMulti-step Reasoning

Often confused with
Long context window
Refers to the amount of text an AI model can process at onc
Batch processing
Involves processing many independent tasks simultaneously, rat
Reinforcement learning
A training method where an agent learns through trial an
Complex reasoning
A broader term that can include long-horizon tasks but also a
Origin

The concept of long-horizon tasks has emerged as a natural progression in AI research, particularly with the development of more capable AI agents and large language models. It describes the challenge of extending AI capabilities beyond immediate responses to sustained, goal-oriented activities.

Try this about 5 minutes

Ask an AI assistant like ChatGPT or Claude to write a short story, then ask it to continue the story with a specific plot twist, and then ask it to summarize the entire story. This multi-turn interaction, especially if the story is long, demonstrates the AI's ability to maintain context and build upon previous steps.

Try it: Playbook entries that use thisPlaybook →

The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.

Questions people ask

What makes a task 'long-horizon'?
A task is considered long-horizon if it requires many sequential steps, takes a significant amount of time to complete, or involves maintaining context and state over an extended interaction.
How do AI agents handle long-horizon tasks?
AI agents use techniques like task decomposition to break down large problems into smaller steps, planning to sequence these steps, and memory to retain information across the task's duration.
Are long-horizon tasks the same as tasks requiring a large context window?
Not exactly. A large context window allows an AI to process more information at once, which is helpful for long-horizon tasks, but the task itself is defined by its duration and multi-step nature, not just the amount of input data.

First seen on Subvolts Dec 11, 2025. Updated Oct 14, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.