Organizations historically treated team size as a primary metric for productivity and organizational impact. Aaron Goldsmid argues this proxy collapsed over the past 18 months, shifting product leadership attention entirely toward individual leverage and execution tooling rather than managing headcount ratios.
Frontier models can be slow, expensive, and prone to hallucinating across non-deterministic training sets. Training a small language model exclusively on a verified proprietary corpus provides sub-second latencies, cuts operational compute costs, and matches or exceeds premier frontier models on domain tasks.
Rather than celebrating generic wins, high-performing engineering teams actively map the precise edges where LLMs fail. Discerning which tasks are computational versus narrative keeps teams from shoehorning tabular or deterministic logic into textual forms that trigger severe signal loss and error propagation.
Deel unlocked substantial hidden revenue and capacity by empowering operations, sales, and partnerships to automate their own workflows instead of bottlenecking on engineering. These granular, manual tasks were previously ignored because handcrafting software was historically too expensive to justify building them.
Functional guilds like engineering, design, and product naturally develop divergent subcultures and viewpoints over time. Forcing line-level cross-functional teams to present dirty laundry together every two weeks limits drift, ensures shared alignment on disagreements, and eliminates conflicting 1:1 side directives.
AI radically reduces the upfront cost of rewriting software, making legacy codebases easy to replace rather than endure ongoing tech debt. The emergent danger is the loss of embedded domain knowledge, as citizen developers repeat classic maintenance mistakes without extracting the nuanced edge-case insights of existing systems.
Traditional automation handles simple, deterministic data movement but breaks when encountering minor format anomalies or routine ambiguity. Large language models succeed by automating the vast operational middle layer—tasks demanding slight context-aware judgment—without needing to replace high-stakes human strategic decisions.