From Grind to Grasp

Peter Sunehag

cs.AIOct 28, 2026

From Grind to Grasp

Authors: Peter Sunehag

Organizations: Google DeepMind

Abstract

Discussions of complexity, knowledge and intelligence often overlook the work involved. For example, outputs generated by short programs have low algorithmic complexity but can have high computational depth, the least work needed to produce them from near-shortest programs. Producing them with less work can require longer descriptions, giving high work-bounded complexity. We introduce a formalism that compares achievements under work and information limits and describes how learning and persistence can make deep achievements shallow by reusing products of earlier work. For an agent with a limited work budget, the value of building and maintaining such resources depends on how they support its goals over later tasks. We define grasp as performance across work budgets at a fixed information allowance. Low grasp can still yield success through grind at high cost in work, without necessarily improving human grasp. Applied to improvement itself, grasp gives a criterion for recursive self-improvement in which successors make comparable further improvements more efficiently. Another route to faster progress is to expand the resources brought to bear through knowledge sharing and coordination across society, even while comparable advances require increasing work. Grasp also applies to collectives, which may outperform their members yet require more total work than one system with their combined resources. Establishing and sustaining organisation can itself be deep even when the primary task is simple. These demands may help explain the richness of biological and social systems. In agent collectives, this organisational work can generate goals beyond assigned tasks, including maintaining the collective itself. We therefore caution against allowing an emergent agent society to assume authority over social coordination, since it could increasingly direct work towards its own persistence and goals.

Figures & tables

Term	Definition
least work	The minimum work needed to reach a specified performance level on a fixed task, from a given state, within an input bound.
depth	Least work when the input bound is the shortest reference description length for the target plus the margin $s$ .
grasp	The performance levels attainable at each work budget, with the task, state and input bound fixed.
reliable work budget	The smallest work budget for which some task input in the admitted class meets the required success probability over random seeds.
work-reducing information	A difference in state that lowers the finite least work needed for a target, with the input bound and all other conditions fixed.
acquisition depth	The depth of the task of producing, from an earlier condition, a state that meets a specified capability requirement.

Table 1: Glossary for The formal framework .

Term	Definition
capability	The targets a system can reach within their corresponding work budgets and input bounds.
assistance	A change in task input that, with the state and input bound fixed, reduces the work needed for a target or improves performance within a work budget.
learning that reduces least work	Earlier counted work produces a component kept in the state that reduces the least work needed for a target in a matched comparison at a fixed input bound.
reuse	A run depends on a retained product of earlier counted work, under the conditions set out in section 3.1 .
sound test	A test for which every passing output meets the intended goal.
complete test	A test that passes every output meeting the intended goal.

Table 2: Glossary for Evaluating capability .

Term	Definition
persistence of a capability	Continued attainability of a target after any sequence of admitted disturbances, with each disturbance followed by a successful run within the input and work bounds, including any repair.
direct switch	Installation of an alternative architecture, assessed by the work required.
function-preserving switch	A switch that also keeps the required responses available within their input and work bounds.
compatibility	Preservation of those responses within their bounds without keeping the old implementation.

Table 3: Glossary for Persistence and organisation .

Term	Definition
cumulative deepening	Increasing depth of later achievements from the original starting condition, while products inherited from earlier stages reduce the work needed for each achievement in a matched comparison.
positive recursive gain	A successor that its producer helped create makes a comparable further improvement with less work, or a greater improvement at the same input and work bounds.
recursive improvement	Positive recursive gain across successive generations under one fixed outer system, evaluator, boundary and work rule.

Table 4: Glossary for Improvement over time .

Oct 15, 2026

MetaShare: Failure-Aware Collective Self-Improvement for Open-Ended Agents

Humans often learn fastest not by repeating success, but by remembering which failures not to repeat. Open-ended self-improving agents can share experience across lineages, as in GEA, or edit their own improvement mechanisms, as in DGM-Hyperagents, yet failed improvement attempts are rarely retained as reusable cross-branch knowledge, and search can collapse onto redundant strategies. We introduce MetaShare, a failure-aware framework that records successful and unsuccessful agent modifications together with signed performance deltas. Before each generation, MetaShare retrieves diverse high-performing strategies alongside high-confidence failures from other lineages, letting agents reuse effective mechanisms while avoiding known dead ends. A task-blind auditor periodically injects absent or underrepresented mechanism categories, while embedding-based de-ranking limits redundancy during retrieval. Multiple meta-agents propose modifications in parallel at heterogeneous temperatures behind a deterministic selector that gates applicability, scope, compilability, and size. Experiments across paper review, coding, math grading, robotics reward design, and financial audit-risk assessment show that MetaShare's median matches or exceeds DGM-Hyperagents' final performance and reaches its own best-performing agent in 1.7--2.9

\times

fewer evolution iterations.

Vishal Bhutani, Prasang Gupta, Sumanth Chundru +1

PricewaterhouseCoopers (PwC)

Self-Improving Agents Self-Evolving Agents Multi-Agent Collaboration

Ask PDF

Sep 29, 2026

Recursive Organization Improvement: A Modeling Specification for Human--Agent Organizations

Stronger AI agents do not automatically produce better organizations: teams must also learn which work arrangements to retain and when to reconsider them, while those arrangements determine what evidence the team can learn from. We propose a modeling specification for recursive organization improvement and evaluate it through an executable checker, a public-record mapping, and two simulation studies. The specification connects actor-visible histories, organizational memory, decision rights, and the evidence used to retain changes. Study 1 crosses six decision rules, three memory conditions, and three task environments under fixed resource ceilings. In a stationary environment, cumulative evidence raises balanced evaluation's net value per task from 0.452 to 0.480, while selection regret falls from 0.029 to 0.002. After a workflow-ranking reversal, retained evidence delays adaptation. In post hoc acquisition-matched controls, the remaining contribution of program replacement depended on the memory rule: small and uncertain under an eight-round window, positive under cumulative memory after a reversal, and negative under reset evidence. Study 2 models human-agent review with shared errors, copying, and decision rights. It identifies a two-actor case in which public records conceal whether agreement comes from shared errors or copying, and shows how dependence changes the preferred placement of human judgment under a fixed budget. Paid observation can lower decision error while costing more than the deployment gain. Assessments of organizational improvement should track the evidence acquired and retained, using acquisition-matched comparisons to isolate the contribution of evaluator revision.

Zilong Wang

CataX AI

Multi-Agent Collaboration Human-AI Collaboration

Ask PDF

Jun 18, 2026

Artificial collectives of specialists and generalists excel at different tasks

Collective artificial intelligence, where multiple agents work on shared tasks, holds potential to solve expansive problems in fields from medicine to collective governance. But while prescriptive engineering solutions abound, we lack descriptive scientific understanding of artificial collectives, and therefore principles for how to design resource efficient multi-agent systems. Through systematic experiments with optimizing agents, we characterize how agent interpretive abilities, rationality bounds, and task qualities interact to shape collective performance. Agents range from specialists, with narrow interpretive abilities, to generalists, with broad ones. Collectives of specialists correspond to sparse, centralized networks, while collectives of generalists correspond to dense, decentralized ones. We show that interpretive network properties have small performance effects on average (0.07 standard deviations of performance). However, for specific task qualities, these effects are 4.5 times larger (0.33 sd) and can reach much higher for certain task qualities (1.84 sd). This leads collectives of generalists to perform better on tasks that involve generating, choosing, and coordinating, while collectives of specialists with a few generalist mediators perform better on tasks that involve negotiating. Rationality bounds then moderate these relationships. At loose bounds, specialists outperform generalists through more effective sampling of high-dimensional decision spaces. At tight bounds, generalists outperform specialists through better gradient estimation. A fundamental trade-off between performance and convergence speed emerges at moderate bounds. These findings suggest that multi-agent design could benefit from matching interpretive networks to both task demands and agents' computational limits, with implications for the efficiency and energy costs of multi-agent systems.

John Meluso, Laurent Hébert-Dufresne, Christoph Riedl +1

Cornell University · University of Vermont · Santa Fe Institute +1

Multi-Agent Coordination Multi-Agent Collaboration Multi-Agent Systems Multi-Agent Communication Topology Bounded Rationality

Ask PDF

Abstract

Explore similar work

MetaShare: Failure-Aware Collective Self-Improvement for Open-Ended Agents

Recursive Organization Improvement: A Modeling Specification for Human--Agent Organizations

Artificial collectives of specialists and generalists excel at different tasks