01
APPARITION
A coding agent comes into existence as a viable engineering actor: it interprets intent, explores repositories, uses tools, modifies software, executes commands and iterates.
The model
From agents to engineering systems
02
SELECTION
WORKING HYPOTHESIS
Engineering environments filter coding agents and their configurations. Coding agents that better fit their environment are more likely to remain useful as scale and autonomy increase. Fit coding agents produce useful outcomes without proportional growth in cost, risk or human attention.
03
COOPERATION
Fit coding agents cooperate when working together improves their ability to succeed or persist compared with operating independently. They compose into complex workflows: work is divided, state is exchanged, capabilities combine and coordination becomes an engineering problem.
04
SPECIALIZATION
Coding agents can develop persistent, differentiated roles through repeated stable cooperation. At that point, the cooperating system itself can become the meaningful engineering unit — a higher-order agentic system.
INDIVIDUAL AGENT1 agentcapability matters
AGENT POPULATION1K agentsenvironment matters
Working hypothesis: agentic engineering is firmly in Selection and may be entering Cooperation.
Selection
The environment becomes part of the fitness function
The useful question is which engineering conditions allow agent variants to operate reliably and economically at scale.
Candidate DevEx conditions for investigation, not an organizational roadmap.
CONTEXTCan agents efficiently acquire sufficient knowledge?
EXECUTIONCan agents safely and reproducibly perform engineering work?
VERIFICATIONCan agents produce trustworthy evidence of correctness?
COORDINATIONCan parallel work compose without coordination cost dominating?
OBSERVABILITYCan humans understand behavior, outcomes and exceptions?
ECONOMICSDo compute, tools, CI and human attention remain viable?
LEARNINGDo failures improve the environment for future executions?
Agentic-system fitness = useful outcome / ( context + execution + verification + coordination + observability + economics + learning )
The forcing function
What breaks at 1K agents?
1K agents is not a utilization target or industry benchmark. It is a stress test: small inefficiencies become systemic when multiplied across an agent population.
Expensive context exploration × 1Kbecomes systemic waste
Non-reproducible execution × 1Kbecomes operational instability
Human-dependent verification × 1Kbecomes a throughput ceiling
Coordination conflicts × 1Kcan erase the benefit of parallelism
Opaque agent behavior × 1Kmakes failures and exceptions impossible to understand
Unbounded execution cost × 1Kturns small inefficiencies into unsustainable economics
Repeated failures × 1Kcompound unless every execution improves the system
1K-agent scale = amplify( friction + ambiguity + risk + coordination + cost )