Elizabeth Ayer on Mastodon — "The fact that we are not seeing wildly improving software all around us tells us everything we need to know"

Elizabeth Ayer · article · 2026-04-10 Reader 🖥️

Code Was Never the Bottleneck Code Quality

The absence of visible improvement in software quality and innovation, despite the rise of AI code generation tools, demonstrates that code creation was never the bottleneck for delivering software value.

practical · social media post (Mastodon) — A brief polemical observation about AI-driven software development, structured as a modus tollens argument. Despite its brevity, it makes a substantive claim about the relationship between code generation and value delivery.

Structure

Part	Role
Opening observation: we are not seeing wildly improving software	Establishes the empirical premise — the expected consequence of AI code generation is not materializing
Elaboration: no flourishing of value delivery, new product categories, or better-satisfied needs — it's the opposite	Strengthens the premise by specifying what 'improvement' would look like and asserting its absence (and reversal)
Conclusion: all we are seeing is decreases in quality, because code creation is not the problem	States the causal explanation — the root cause of software shortcomings lies elsewhere than code production

Key terms

value delivery Code Was Never the Bottleneck
The end-to-end process of identifying human needs and shipping software that satisfies them — not just writing code, but the full chain from understanding to impact
code creation Code Was Never the Bottleneck
The act of producing source code, which AI tools like LLM-based assistants have dramatically accelerated
the problem
The binding constraint on software quality and innovation — the actual bottleneck preventing better outcomes
quality Code Quality
The degree to which software meets real user needs reliably, usably, and durably — not merely 'works' or 'has features'

Propositions

explicitWe are not seeing wildly improving software despite AI code generation capabilities.conf: high
explicitThere is no flourishing of value delivery, new product categories, or better satisfaction of needs.conf: high
explicitSoftware quality is actually declining.conf: high
explicitdevelopedCode creation is not the bottleneck for software quality or innovation.conf: high
extends ←How Leaders Can Use AI to Solve Real Business Problems · p7 — Ayer's claim that code creation isn't the bottleneck gets a mechanism here: faster code production (Claude Code) actively removes the friction that used to force alignment with the real brief, making the non-code work more, not less, load-bearing.
refines ←Presentation: Leadership in AI-Assisted Engineering · p6 — Agrees code generation is not the bottleneck and names the actual time sinks: interruptions, context switching, and meetings.
implicitIf code creation were the bottleneck, AI-assisted code generation would have produced visible improvements in software outcomes.conf: high
implicitcorroboratedThe real bottleneck lies in activities upstream or downstream of code creation — understanding needs, design, integration, maintenance, organizational alignment, or similar.conf: medium
supports ←Presentation: The Future of Engineering: Mindsets That Matter When Code Isn't Enough · p10 — Both locate engineers' durable value upstream and downstream of code creation — in understanding the customer's actual problem — rather than in the act of producing code.
implicitAI code generation tools may be actively contributing to quality decline by making it easier to produce code without addressing the actual constraints.conf: medium

Arguments

Code creation is not the bottleneck for delivering software value. strength: high

Explicit premises:

AI has dramatically increased code creation capacity.
We are not seeing improvements in software quality, value delivery, or innovation.
Software quality is actually decreasing.

Implicit premises:

If code creation were the binding constraint, relaxing it (via AI) would produce visible improvements in outcomes.
The absence of improvement when a supposed constraint is relaxed is strong evidence that the constraint was misidentified.
AI code generation tools have been widely enough adopted to test this hypothesis at scale.

Inferences:

Modus tollens: If code-creation-is-the-bottleneck then we-would-see-improvement. We do not see improvement. Therefore code-creation-is-not-the-bottleneck.
The decline in quality suggests that accelerating code production without addressing the real bottleneck may be counterproductive.

Narratives

The implicit narrative of the AI hype cycle: the tech industry expected that making code cheaper to produce would unlock a golden age of software innovation, but the golden age has not arrived.
→ Sets up the expectation that the argument then subverts — the reader is assumed to be familiar with the AI productivity promise

Questions addressed

✅ Is AI code generation making software better?
No — there is no visible improvement, and quality is declining.
⭕ What is the real bottleneck for software quality and innovation?
The post identifies what it is NOT (code creation) but does not specify what it IS.

Gaps

unstated alternativeThe post argues code creation is not the problem but never identifies what the actual bottleneck is. Candidates include: understanding user needs, product design, organizational incentives, maintenance burden, regulatory friction, distribution, or the economics of attention.
Read next: Read Adler's 'syntopical' framing: compare with Theory of Constraints literature (Goldratt), Marty Cagan on product discovery, and Rich Hickey's 'Simple Made Easy' on complexity as the real constraint.
closed byIs AI coding on average 17.3x faster or 1.3x faster? Both. — LinkedIn post by Fabrice Bernhard — Names the actual bottlenecks Ayer leaves unidentified: unclear business needs and organisational dependencies, which AI-accelerated code production cannot address (p4, p8).
empirical challengeThe claim that software quality is declining is stated as self-evident but not evidenced. Quality decline in consumer software (enshittification) may coexist with genuine improvement in developer tooling, infrastructure, and specialized domains.
Read next: Compare with Cory Doctorow's enshittification thesis and Ben Thompson's aggregation theory for structural explanations of why consumer-facing software quality declines even as technical capability increases.
scope ambiguityIt is unclear whether 'code creation is not the problem' means (a) it was never a significant constraint, or (b) it was a constraint but is now solved and the next constraint is revealed. These have different implications for whether AI tooling is still valuable.
Read next: Theory of Constraints: when you relieve one bottleneck, the constraint moves. The post may be observing a constraint shift rather than proving code creation was never relevant.
closed byThe keyboard was never the bottleneck — Resolves Ayer's scope ambiguity with an articulated position: code production was never the constraint (p1), and AI's effect is to compress the lower levels while exposing the higher-level constraints (p3), i.e. both never-was and constraint-shift are true at different levels.
selection biasThe post does not address survivorship bias: AI-augmented software improvements may be happening in less visible domains (internal tools, infrastructure, niche products) while the most visible consumer software degrades for unrelated economic reasons.
Read next: Look for data on AI-assisted development outcomes in enterprise/internal tooling vs. consumer products.

Elizabeth Ayer on Mastodon — "The fact that we are *not* seeing wildly improving software all around us tells us everything we need to know"