DEV Community: Samantha Lambert

How AI Is Rewriting the Core Abstractions of Software Engineering

Samantha Lambert — Tue, 28 Apr 2026 07:37:50 +0000

A common question we all have is, "Will AI replace developers?"

In practice, that’s not what’s happening. What’s changing is far more subtle and far more important. The fundamental unit of work in software engineering is shifting.

For decades, engineers were valued for their ability to translate requirements into precise, deterministic code. Today, that boundary is blurring.

We are moving from a world where code is the primary artifact to one where context (the constraints, intent, and structure surrounding a system) matters more than the code itself. AI isn’t removing the need for engineering; it is redefining what engineering actually is.

Software Engineering Before AI
It was fundamentally about control. Every behavior a system exhibited had to be explicitly defined, implemented and maintained by a human.

The core abstractions, including functions, classes, modules, and APIs were all designed to organize that control into manageable units.

Back then, developers were translators. Business requirements were converted into logic, edge cases were handled through conditionals, and correctness was achieved through exhaustive specification.

If something worked, it was because someone had written the exact instructions to make it work. If it failed, the failure could usually be traced back to a specific line of code.

It created a predictable system: deterministic inputs produced deterministic outputs. Testing followed the same philosophy; verify that given inputs lead to expected results. Engineering, at its core, was the discipline of reducing ambiguity through explicit implementation.

How Abstraction Layers Changed Everything
AI is often described as a disruption, but software engineering has always evolved through abstraction. Moving from assembly to high-level languages removed the need to manage memory directly. Frameworks abstracted away repetitive patterns. Cloud platforms eliminated the need to think about physical infrastructure.

Each step moved engineers further away from low-level concerns and closer to intent. Instead of managing servers, we define services. Instead of writing boilerplate, we compose systems.

AI is the next step in that progression. It doesn’t replace abstraction; it accelerates it. But unlike previous shifts, it doesn’t just abstract implementation details. It begins to abstract the act of writing code itself.

AI Changes the Unit of Work: From Code to Context
With AI systems, code is no longer the primary bottleneck. You can generate functions, scaffolding, and even entire modules in minutes. What becomes scarce instead is clarity: clarity of intent, clarity of constraints, and clarity of expected behavior.
The new unit of work is context.

This includes:
• Defining what the system should do (intent)
• Constraining how it behaves (rules and boundaries)
• Evaluating whether the output is acceptable (validation)

As Andrej Karpathy described with the concept of “Software 2.0,” systems are increasingly described rather than fully written. Instead of specifying every step, engineers guide systems toward outcomes.

In this model, writing code becomes just one part of a broader responsibility: shaping how a system behaves under varying conditions. The engineer’s role shifts from builder to orchestrator.

What Actually Changes Technically
The shift to AI-assisted development doesn’t just change how code is written; it changes how systems behave, fail, and are maintained. Many of the familiar engineering activities still exist, but their nature is evolving in subtle ways.

Tasks that were once deterministic and code-centric are becoming more about interpreting behavior, managing uncertainty, and maintaining consistency across systems that are no longer entirely hand-written.

Understanding these changes at a technical level is key to adapting effectively, because the tools may be new, but the responsibility for system reliability still sits with the engineer.

Debugging Becomes Behavior Analysis
In traditional systems, debugging meant tracing execution paths and identifying faulty logic. With AI-generated or AI-assisted systems, the failure modes are different. The code may be syntactically correct, yet produce unexpected or inconsistent outputs.

Debugging becomes less about fixing a line of code and more about understanding why the system behaved the way it did. This includes analyzing prompts, inputs and environmental context.

Code Ownership Becomes System Ownership
When AI generates significant portions of code, the idea of ownership changes. You may not have written every line, but you are still accountable for the system’s behavior.

Ownership shifts upward from implementation details to system-level reliability, performance, and correctness.

Testing Becomes Probabilistic
Traditional testing relies on deterministic assertions. But AI-driven components introduce variability. The same input may yield slightly different outputs, all of which could be acceptable or problematic.

Testing evolves into evaluating output quality, confidence, and consistency rather than simple pass/fail conditions. Systems must be designed to accommodate changing behaviors and feedback loops, not just static correctness.

What Doesn’t Change
Despite these shifts, the core responsibilities of software engineering remain intact.

Systems still need to be designed. Trade-offs still need to be made. Constraints still need to be defined. AI can generate solutions, but it cannot determine which solution is appropriate in a given business or technical context.

It does not understand long-term maintainability, regulatory implications or the subtle interactions between system components. Those remain firmly in the domain of human judgment.

A useful way to think about it is this: AI can generate solutions, but it cannot define what “correct” means. That definition and the responsibility that comes with it still belongs to engineers.

The New Skill Stack for Engineers
As the abstraction layer shifts, so does the required skill set. The emphasis moves away from syntax mastery and toward:
• Context engineering: Defining clear inputs and constraint
• System thinking: Understanding interactions across components
• Evaluation skills: Judging output quality and reliability
• Communication: Articulating intent clearly to both humans and machines

This doesn’t mean traditional skills disappear, but they are no longer the differentiator. The ability to reason about systems at a higher level becomes more valuable than the ability to write code quickly.

The Hidden Risk You Should Know
One of the least discussed consequences of AI-assisted development is the acceleration of system entropy.

When code can be generated quickly, it tends to be generated more often. This increases volume, but not necessarily coherence. Without strong architectural discipline, systems can become fragmented, inconsistent, and difficult to maintain.

AI can produce working code, but it does not inherently enforce consistency across a codebase. Over time, this leads to diverging patterns, redundant logic, and reduced clarity.

In other words, the speed gains from AI can be offset by long-term maintenance costs if systems are not carefully managed.

Engineering Isn’t Dying, It’s Moving Up
Software engineering is not being replaced. It is being transformed.

The focus is shifting from writing code to shaping systems. Engineers are becoming less like typists and more like system designers, responsible for guiding increasingly complex, semi-autonomous systems.

The best engineers in this new era won’t be the ones who write code the fastest. They’ll be the ones who can define problems clearly, constrain solutions effectively, and reason about systems in a world where code is no longer the limiting factor.

DePIN Explained: How Web3 Is Building Real-World Infrastructure

Samantha Lambert — Mon, 27 Apr 2026 08:37:45 +0000

For years, Web3 lived mostly on screens. Tokens, NFTs, DeFi platforms, all digital. DePIN changes that. Short for Decentralized Physical Infrastructure Networks, it’s a model that connects blockchain incentives to real-world infrastructure.

Think wireless networks, storage systems, and sensor networks built and operated by individuals instead of corporations. It’s one of the few areas where Web3 is directly interacting with the physical world.

How DePIN Actually Works
The idea is straightforward. Instead of a company building and owning infrastructure, individuals contribute resources. That could be internet hotspots, storage space, or hardware devices. In return, they earn tokens. This creates a decentralized network where participation is open and incentivized.

"The DePIN sector is fundamentally different because it provides real-world utility to end users. Success shows up first in usage and cash flow, not in speculative price action," says Markus Levin, the Co-Founder of XYO

Projects like Helium have already demonstrated this model, with hundreds of thousands of nodes deployed globally at its peak.

Why DePIN Is Gaining Momentum
There are two main reasons: cost and access. Traditional infrastructure is expensive to build and maintain. DePIN distributes that cost across participants, which can make expansion faster and more efficient.

As Riccardo Pagano, the Accelerator Lead at Outlier Ventures, puts it, "DePIN is set to become the backbone of a future where machines and devices transact autonomously. We are mature enough to disrupt massive industries and agile enough to reshape markets.”

It also opens opportunities in underserved areas. Instead of waiting for large telecom companies, communities can build their own networks. According to Messari, a crypto research firm, the DePIN sector could grow into a multi-hundred-billion-dollar market over time.

What DO Industry Leaders Have to Say About It?
The concept is gaining attention from serious builders. Chris Dixon from Andreessen Horowitz has emphasized that “the next phase of crypto is about real-world infrastructure, not just financial speculation.”

Chris Dixon, Founder and Managing Partner at a16z crypto is shocked. He says,“I've never seen a situation in technology where the gap between what I believe is the potential of the technology and the perception is so wide.”

His perspective reflects a broader shift in priorities across the industry.

Here are Some Real-World Applications
DePIN is already being applied in several areas:

Wireless connectivity networks
Decentralized cloud storage
Environmental and IoT sensor networks

“The real-world advantage of DePIN is cost efficiency. By removing the overhead of centralized middlemen and leveraging crowdsourced hardware, these networks can offer services at significantly lower rates.” Staff Insights, Analysis Team at VALR.

These are not experimental ideas. They’re functioning systems, though still early in scale.

Challenges That Can’t Be Ignored
Despite the potential, DePIN faces real obstacles. Scaling physical infrastructure is harder than scaling software. Hardware costs, maintenance, and reliability all come into play.

There’s also regulatory complexity, especially in sectors like telecom. And quality control remains a concern. Decentralization can introduce variability that centralized systems avoid.

Why It Matters Long Term
DePIN represents something Web3 has struggled with: tangible utility. It connects blockchain incentives to real economic activity. That makes it easier to explain, easier to justify, and potentially more sustainable.

It also aligns with global trends toward decentralization and resilience in infrastructure.

The Bridge Between Two Worlds
DePIN is not the most talked-about part of Web3, but it may be one of the most important. It moves the conversation from digital speculation to physical systems. If it succeeds, it could redefine how infrastructure is built and owned. That’s a much bigger shift than any token cycle.

AI Meets Web3: Is it a Real Shift or Just Another Tech Hype Cycle?

Samantha Lambert — Fri, 24 Apr 2026 14:27:59 +0000

If you spend even a few minutes in tech circles right now, two terms dominate the conversation: AI and Web3. Individually, they’ve each gone through waves of hype. Together, they’re being framed as the next big leap.

But, there’s a real question underneath the noise. Is this a meaningful convergence, or just another moment where two buzzwords collide without substance?

The answer sits somewhere in between, but it’s leaning toward something real. As Werner Vogels, CTO of Amazon Web Services has often noted, the evolution of enterprise infrastructure relies on the shift toward distributed systems that handle large-scale computing with high efficiency. This transition is now being fueled by the need for more transparent AI.

Where Does AI and Web3 Intersect?
At a practical level, the overlap comes down to three things: data, ownership, and trust. AI systems depend heavily on data; the more they have, the better they perform. But that creates a problem: most of that data is controlled by large centralized platforms. Web3 introduces a different model, allowing individuals to own and control their data through blockchain-based identities and wallets.

As Vitalik Buterin, co-founder of Ethereum, put it in a 2024 blog post, “AI and blockchain are complementary in that blockchain can provide credible neutrality and transparency for AI systems”. This idea is driving much of the experimentation happening now.

Roman Milyushkevich, CTO of HasData, has seen this firsthand in adaptive data markets. He points out that Web3 is moving beyond "ownership hype" toward verifiable data provenance.

In these systems, every piece of data has a cryptographic trail, allowing AI models to dynamically weight inputs based on trust history rather than just accuracy. This creates a self-adjusting model capability that traditional, implicit pipelines struggle to enforce.

Here are Some Real Use Cases
Some of the most credible use cases are already in progress. One is decentralized AI marketplaces, where developers can access datasets and models without relying on a single provider. Another is the creation of tamper-evident audit trails for AI-generated content.

Jay Speakman, CTO at CustomWritings, highlights the importance of "on-chain auditability" for AI outputs. By hashing every prompt and model version on a ledger, teams can reconstruct and verify exactly how a specific output was generated.

"The real shift here is using Web3 as a trust layer for inherently probabilistic AI systems, something traditional infrastructure can’t fully guarantee," Speakman explains.

Humayun Sheikh, CEO and Founder of Fetch.ai, is taking this a step further by building decentralized platforms where autonomous AI agents can search for data, negotiate deals, and share insights without intermediaries. This shift aims to break data monopolies and return power to individuals through permissionless networks.

Building on this, Ankush Verma, CTO of EssayShark, sees a major opportunity in verifiable decision-making. When AI is used to grade essays or rank applicants, users want to know how it arrived at its conclusion. By anchoring cryptographic proofs on-chain, platforms don't have to ask for blind trust; they offer a way for users to verify that results haven't been altered.

Determinism vs. Probabilism
Despite the potential, integrating these two technologies is not seamless. Andrew Libby, co-founder and CTO at StatusGator, points out a fundamental architectural friction.

"AI is probabilistic, while blockchain is deterministic. If the code doesn't match exactly on a blockchain, the block fails. Because of the high compute costs, most teams are forced to do the heavy AI lifting off-chain and only push the final results back to the ledger. This often leads to a "verification lag" that can kill the real-time speed users expect."

This challenge is echoed by John Wiegley, CTO of Kadena, who applies mathematical principles like category theory to enhance engineering design in blockchain. He continues to drive innovation by seeking ways to merge deep technical rigor with the need for scalable, decentralized AI systems.

Milyushkevich notes that teams are solving this by decoupling execution from verification. They use a lightweight on-chain layer to verify "proofs of process" rather than the full computation. Techniques like zero-knowledge proofs (zk-proofs) are increasingly used for specific model steps, though they remain expensive and limited in scope.

As Marek Olszewski, co-founder of Celo , emphasizes, the goal is to make these complex decentralized systems accessible and mobile-first, ensuring that even DeFi and AI-driven identity are available to anyone with a smartphone.

Key Statistics You Should Know
The numbers show why this space is attracting attention:
• The global AI market is projected to add $15.7 trillion to the economy by 2030 (PwC estimates).
• Venture funding for AI-related crypto projects has increased significantly since 2023.
• By 2026, roughly 70% of requests for quotes are expected to be handled by AI agents without human involvement.
• The AI crypto market capitalization surpassed $26 billion in early 2026, reflecting massive developer activity.

Risks and Open Questions
There are still major uncertainties. Data privacy remains a concern; making data more accessible through Web3 can sometimes conflict with the need to protect it. There’s also the risk of "centralization creeping back in," especially as AI models scale and require more massive resources.

Regulation is another unknown. Governments are still figuring out how to handle both AI and blockchain separately, let alone their combination. David Schwartz, CTO of Ripple, continues to work on cross-border financial infrastructure while navigating these evolving regulatory and scalability questions.

Take Away: Maybe Promising, but Early
AI and Web3 are not a perfect match, but they solve complementary problems. The combination has real potential, especially around data ownership and transparency. As Libby notes, the difference lies in having a verifiable, audited history rather than just trusting a vendor's word.

At the same time, technical and practical limitations mean it’s still early. This isn’t just hype, but it’s not fully formed either. What matters now is whether builders—like those at HasData, CustomWritings, StatusGator, and EssayShark—can move beyond concepts and deliver systems people actually use.

Did You Know About the Battle to Make Ethereum Usable?

Samantha Lambert — Thu, 23 Apr 2026 13:51:39 +0000

Ethereum has long been the backbone of Web3. It supports everything from decentralized finance to NFTs and countless applications in between. But there’s a problem that hasn’t gone away.

When demand increases, the network becomes expensive and slow. Fees rise, transactions take longer, and basic interactions can become frustrating. For something positioned as the future of digital infrastructure, that’s a serious limitation.

Vitalik Buterin, Co-founder of Ethereum, has noted that while the network is robust, "improving usability is more than a technical challenge. It is an opportunity to enhance the user experience."

Solving this issue has become one of the most important challenges in Web3 today.

Why Scalability Is Such a Big Deal
At its core, Ethereum only processes a limited number of transactions per second. That constraint creates congestion when usage spikes. The consequences are immediate: users pay higher fees, smaller transactions become impractical, and developers are forced to design around these limitations.

Sandeep Nailwal, Co-founder of Polygon, describes this as a visceral experience rather than a theoretical one. He points out that the scalability hurdle is felt "every time a transaction costs more than lunch and the network slows."

If blockchain is going to support millions of users, this model doesn't hold. That’s why scalability is not just a technical issue—it’s a usability issue.

The Rise of Layer 2 Solutions
Instead of replacing Ethereum, the current approach is to build on top of it. Layer 2 (L2) networks handle transactions separately and then settle the results back on the main chain. This reduces congestion while still benefiting from Ethereum’s security.

Steven Goldfeder, CEO and Co-founder of Offchain Labs (the team behind Arbitrum), explains the relationship simply: "We take as much computation off Ethereum as possible while having Ethereum be the referee. If something goes wrong, Ethereum can fix things." The goal is universal: make transactions faster and cheaper without compromising trust.

A Competitive and Fragmented Landscape
What makes this moment interesting is the level of competition. Multiple Layer 2 networks are trying to solve the same problem in slightly different ways. Some prioritize efficiency, others focus on security or decentralization.

This competition is driving rapid innovation, but it’s also creating fragmentation. Users and liquidity are spread across different ecosystems, which adds complexity. For someone entering the space, the experience can feel confusing. Choosing where to transact or build is no longer straightforward.

What the Data Shows
Usage trends suggest that Layer 2 solutions are not just experimental. Platforms tracking blockchain activity show that a significant portion of transactions now happens off the main Ethereum chain.

In some cases, Layer 2 networks collectively handle more activity than Ethereum itself. This shift shows that users are willing to move if the experience improves.

What Developers Are Prioritizing
Developers are increasingly building on Layer 2 networks for a simple reason: cost. High transaction fees limit user growth. Jesse Pollak, Creator of Base and Head of Protocols at Coinbase, emphasizes that the mission is to "bring the next billion users to web3."

He argues that "the only thing that matters is delivering products to customers that they love," and high friction costs break that experience before it even starts.

The Trade-Offs No One Can Ignore
While Layer 2 solutions solve some problems, they introduce others. Security is one concern. Even though these networks rely on Ethereum, they add additional layers that need to be secure on their own.

Another issue is interoperability. Moving assets between different networks can be complex and sometimes risky. Ben Jones, Co-founder of Optimism, envisions a future where these chains "talk to each other and become the new backbone for the internet."

He further admits that getting there requires moving from a system of unique, siloed chains to a consistent, "sharded" environment that feels like one single network to the user.

Will There Be a Clear Winner?
One open question is whether a single solution will dominate. It’s possible that different Layer 2 networks will specialize in different use cases. Some may focus on finance, others on gaming or social applications.

This would lead to a more diverse but also more complex ecosystem where users might interact with multiple networks without always realizing it.

What This Means for the Future of Web3
If scalability solutions continue to improve, the impact will be significant. Transactions could become fast and inexpensive, removing one of the biggest barriers to adoption.

Developers would focus less on infrastructure constraints and more on user experience. At that point, the conversation shifts. Instead of asking whether blockchain works, people start asking what they can build with it.

Progress, Not the Finish Line
The effort to scale Ethereum is one of the most important developments in Web3 right now. It’s not as visible as market trends or token launches, but it underpins everything else. Without it, broader adoption remains limited.

Progress is being made, and the direction is clear. Ethereum is evolving into a foundation layer, supported by a growing ecosystem of scaling solutions. The real test is whether this delivers a seamless experience for everyday users. If it does, it could finally unlock the potential that Web3 has been promising for years.

Why Web3 Is Quietly Rebuilding Around Real Value

Samantha Lambert — Wed, 22 Apr 2026 11:16:24 +0000

A couple of years ago, Web3 felt loud. Everywhere you looked, there were new tokens, NFT drops, and bold claims about reinventing the internet. Then the market cooled, and much of that excitement disappeared almost overnight.

What replaced it is easier to miss, but far more important.
Behind the scenes, builders have been shifting focus. The conversation is no longer centered on hype cycles. It’s about whether blockchain can connect to real economic activity in a way that actually works. That shift is subtle, but it marks a turning point for the industry.

What “Real Value” Means in Today’s Web3

The current phase of Web3 is less about abstract digital assets and more about tying blockchain systems to things people already understand. Think financial instruments, ownership rights, and income-generating assets.

Instead of creating value from speculation alone, projects are trying to anchor digital tokens to something tangible. This includes assets like bonds, private credit, and even property. The idea is simple: if blockchain is going to matter long term, it has to plug into the real economy.

This is often described as tokenization, even though the broader concept is credibility. Users are starting to care less about what’s possible and more about what’s reliable.

Why This Shift Is Happening Now
Part of the answer is fatigue. After multiple boom-and-bust cycles, both users and investors are more cautious. The question “what’s backing this?” comes up far more often than it used to.

There’s also growing institutional interest. Large financial players are experimenting with blockchain not as a replacement for existing systems, but as a way to improve them. Reports from firms like Boston Consulting Group suggest tokenized assets could become a multi-trillion-dollar market over time.

That kind of projection changes the tone of the conversation. It signals that this isn’t just a niche experiment anymore.

What is Driving Adoption?
Here are the reasons why Web 3 adoption is gaining traction.
•Accessibility- blockchain allows assets to be divided into smaller units, which means more people can participate. You don’t need large capital to gain exposure to assets that were previously out of reach.

•Liquidity- assets that traditionally take time to buy or sell can move more easily in tokenized form. That flexibility is attractive, especially in markets that are otherwise slow.

•Transparency- transactions recorded on a blockchain create an auditable trail. While this doesn’t eliminate risk, it does reduce reliance on opaque systems.

These advantages speak to the inefficiencies that currently exist.

What Builders Are Saying
Across developer communities, there’s a noticeable shift in tone. The focus is less on launching the next viral token and more on building systems that can survive long-term use.

A common sentiment you’ll hear in conferences and forums is that the industry is “growing up.” Builders are prioritizing infrastructure, compliance, and usability over rapid hype-driven growth. In practical terms, that means fewer flashy launches and more work happening quietly in the background.

Challenges to Expect Ahead
Despite the progress, there are two serious challenges that make adopting steady rather than explosive. Regulation is the major challenge. When digital tokens represent real-world assets, legal clarity becomes essential. Ownership has to be enforceable beyond the blockchain itself. Without that, the system lacks credibility.

There’s also the issue of trust between the physical and digital layers. A blockchain can verify transactions, but it cannot independently verify whether an underlying asset exists or is managed correctly. That gap needs reliable intermediaries or new verification systems.

Why This Matters Beyond Crypto
This shift has implications beyond Web3 itself. In regions where access to traditional financial systems is limited, tokenized assets could open new opportunities. Users will no longer need to rely on local infrastructure as they will get the chance to connect directly to global markets. That is why Web 3 is gaining attention in emerging economies. The need for access.

Do We Need a More Grounded Version of Web3?

What’s happening now isn’t as loud as previous cycles. Still, I believe it is more meaningful. Web3 is starting to move away from pure speculation and toward systems that interact with the real world.

That transition won’t happen overnight, and it won’t be perfect. But it represents a necessary step if the technology is going to last.

If this phase succeeds, Web3 won’t be defined by its most chaotic moments. It will be defined by whether it managed to build something useful. What do you think?