Tommy MacWilliam

I’ve been thinking a lot about why products that are technically well-designed can still feel stressful, confusing, or off. I just published a new piece exploring that question through an architectural lens, and would love to hear your thoughts! Read on Substack: https://lnkd.in/eN2jZNBq This post introduces ARCH, a framework I’ve been developing to help teams reason about product coherence, unintended effects, and what products quietly teach over time. Drop a line to let me know what resonates -- or doesn't!

34

Anthropic's Claude Sonnet 4.5 is coming to Glean this week. In addition to being the leading model for coding, our evals show that the model delivers significant improvements over its predecessor by efficiently solving 10% of previously unanswerable enterprise queries, thanks to optimized parallel tool calling and fewer planning iterations. See more details on our blog: https://lnkd.in/gXUEvGKW

28

Design is changing fast. At Indeed, UXers are leading efforts to codify the context of design, content, and code so LLMs can effectively represent our design language. When we identified gaps in prototyping tooling, we took the initiative to build a better way. We're out in front, shaping how the company designs and builds with AI in a way that still feels uniquely "UX". Jenny Wen, who leads design at Anthropic, describes similar shifts on a recent Lenny podcast: the traditional design loop breaking down, design splitting into execution support and short-range visioning, work moving closer to code. In both cases, designers are solving specific problems within their organizations rather than adopting a universal new design approach. This is where UXers shine—people who understand systems, human behavior, and can identify patterns are well-positioned to understand what's needed and drive these discussions. That's not just true for UX. Subject matter expertise in your field positions you to identify where AI can help solve real problems—if you're willing to participate. Across disciplines, people actively shaping this change are having a fundamentally different experience than those waiting to see what happens. Matt Shumer's piece about how everything is about to change has been a topic of discussion over the past few weeks. It's urgent and alarming. He's right that the pace is accelerating faster than most realize. But there's something critical missing from his message: the future is unwritten, and we have agency to shape it by participating. Jenny Wen noted that early-career designers, without legacy processes or deeply ingrained ways of working, are often more adaptable and open to figuring out new approaches. This isn't solely something new designers can embrace. There's a divide: people who feel AI is happening to them tend to lose hope and motivation, clinging to how things are "meant" to be done. Meanwhile, people letting go of how they used to work, imagining something different, and embracing the ride with intention and curiosity are thriving in their work and emotional state. Participation leads to success, which creates energy, which fuels more participation. A virtuous cycle. What does participation look like? - Being curious and open—going with the flow while determining where you're uniquely positioned to add value - Not being afraid to take up space and lead where you might not have led before - Running experiments with peers, sharing what does and doesn't work - Staying oriented on problems and user needs, acknowledging the path to solutions might look different - Playing in adjacent disciplines - Exploring ways AI solves things important to your work, not forcing it where it doesn't make sense Matt's right that this is urgent. But the answer isn't just preparation—it's active engagement. We're co-authors of what comes next. What are you seeing? How are you participating?

46

8 Comments

Figma files for an IPO, plans to trade on the NYSE under the ticker FIG, and reports Q1 revenue up 46% YoY to $228.2M and net income up 233% YoY to $44.9M 

22

1 Comment

Exceptional dinner last night with engineering leaders from a mix of public and late-stage private unicorns. We braved the cold, rainy NYC weather to talk about AI's impact on software development teams. Some highlights from the discussion: - AI Code Gen is like a 24/7 hackathon. Lots of new ideas and prototypes being cheaply created, but the hard part is still making them production-ready. - Even with SDD/TDD, it's impossible to define all the requirements and edge cases in natural language. Your codebase is the more explicit definition of those requirements, and acts as the memory layer for LLMs → so code quality remains important. Especially because LLMs are good at repeating patterns, you want them repeating *good* patterns. - Code review is the new bottleneck, and an increasingly important skill for senior engineers. But it's also an unpleasant job, and can put senior engineers in a reactive position. How might we shift left the guardrails around AI development to mitigate the review burden? - In addition to code review, a big limiting factor is the ability to distribute products to customers effectively. Products needs to be sold, adopted, and used by customers to have impact, and more code gen doesn't solve those problems. - Despite higher velocity of code gen, it's questionable if teams are actually net more productive, in part because of the non-coding limiting factors. It's a good reminder that while code generation has had its big adoption moment in 2025, there's still a lot to figure out in 2026 to truly accelerate development.

101

3 Comments

I’ve come to believe that early experiences carry disproportionate weight - far more than we often acknowledge. For designers and builders, “minimum viable” may suffice for validating core functionality, but it rarely earns trust or sparks genuine delight. While velocity matters, I’ve always struggled with the notion that early versions should be merely viable, especially in a world where attention is scarce and emotional resonance is everything. This extends beyond product. Do you show up as the minimum viable version of yourself in your work, your relationships, your craft? I would encourage you to challenge the conventional MVP framework. Replace it with the Minimum Awesome Product. This doesn’t mean it’s fully-featured or pixel perfect. But it should be worthy of attention. Memorable, even evocative in the way it triggered emotion. The kind of product that prompts someone to say, “I didn’t know I needed this, but now I can’t live without it.” You can refine mechanics over time, but the emotional experience of a first interaction is singular and unrecoverable. Don’t settle for viability. Build something worth remembering… and make it awesome! ❤️‍🔥

148

26 Comments

Taste, Privacy-by-Design & AI-Native Product Teams I’ve been noodling on where privacy work actually might land now that Cursor-style IDEs can spin up a PRD, design doc, and draft pull-request before you finish your coffee. Faster loops aren’t the enemy (even though that's all of our first reactions)—thought-free loops are. Here are three places I see privacy pros adding real value without slowing the train: Sit in the design sprint, not the review queue. AI-generated mocks hit Figma in minutes—jump in that same hour. Ask: Does this risk control feel natural or bolted on? Fix the pattern before it hardens into code. Raise the bar for your toolchain. Adopt better privacy toolchains. Push vendors toward policy-as-code, field-level lineage, and region toggles so AI agents can wire 90 % of controls automatically. Your time is reserved for the edge cases that need human judgment. Point AI at your own work—red-team style. Let models stress-test consent flows, purpose limits, retention rules. Security teams already do this for exploits; direct the cannon at data-use risks, too. Caveat: retrofitting these habits onto a legacy “spaghetti” stack is tougher than starting greenfield—but I think the principles will hold. I'm curious what else your teams are doing today to keep privacy in the loop while the build cycle keeps shrinking? 

34

We’ve been hearing more tension lately between designers × engineers — around speed, hand‑off, and visual fidelity. As roles collapse and tools converge, design engineering is breaking out as its own craft. Not just implementation, but bridging product thinking + code + interaction polish. Why now? 🟢 7500+ open “Design Engineer” jobs in the U.S. today — the momentum around this role is real. (https://lnkd.in/gNhRBXUM) 🟢 23 % employment growth projected for design‑technologist‑style roles this decade (U.S. BLS) (https://lnkd.in/gQ8X7vQH) 🟢 Yet 52 % of teams still use no design‑to‑code tooling, and 73 % say their hand‑off workflow isn’t “good enough” (Telerik State of Designer‑Developer Collaboration 2024) (https://lnkd.in/gwgZSXqw) 🟢 Top‑end salaries show the premium: Netflix’s remote Senior Design Technologist role (Games Tech Research Group) lists $100 k – $720 k (https://lnkd.in/gdC9xXCg) 🟢 Leading-edge companies like Vercel are championing the rise of this role to increase both shipping velocity and design fidelity. (Their guide is excellent → https://lnkd.in/gu3wYjpy) So we’re thinking about an advanced course for the hybrids. What would you want covered? - Front‑end frameworks & component patterns? - Tokens + theming? - Figma‑to‑code & dev‑mode power‑moves? - Rapid prototyping beyond Figma? Drop your must‑have topics below 👇 

26

22 Comments

Wrapping up our Launch Week Commit at Flexprice. Building billing infrastructure isn’t glamorous. It’s one of those things teams postpone, simplify, or patch together… Until product complexity catches up. This week, we shipped the pieces that usually slow teams down: - New ways to aggregate billable events: Count Unique, Latest Value, and fully Custom logic - The ability to apply credits directly to invoices, no manual reconciliation - Calendar-aligned billing : choose between anniversary cycles or fixed calendar dates, so finance teams aren’t left guessing Billing infra should just work, so your team can focus on building product and not fighting spreadsheets or chasing API gaps. It took weeks of iteration, feedback from early teams, and our share of dead ends to get this right. If your pricing is evolving faster than your billing stack, talk to us. We’re building Flexprice as the open-source billing infrastructure for AI & Agentic based companies.

32

5 Comments

Woah, LinkedIn is filled with YC and vibe coding. Why are they pushing it so much? Turns out, it's not just hype, it's capital 😉 they haven't launched a "vibe fund", but they’ve quietly backed dozens of AI-native startups through their standard $500k checks. That’s tens of millions indirectly fueling vibe-coded teams. Nearly 25% of startups in recent batches are building with 95%+ of their code written by AI. Teams of 10 doing the work of 50. Some pulling $1M–$10M ARR. Solar (Lumenary) is one example. But it's bigger than one startup, it’s a pattern. Vibe coding isn’t a trend. It’s becoming a YC lens on the future of product velocity. Still early. But the signal is strong.

19

4 Comments

After managing engineers for years, I discovered something ironic: Team leads are crushing it with AI while senior ICs struggle. Why? Managing people taught them exactly what LLM needs: Breaking projects into clear subtasks Explaining context without assumptions Giving specific, actionable feedback Iterating based on results The delegation skills you built for humans? They're your AI superpower. The best engineers might be prompt-engineering wrong because they never learned to delegate.

42

3 Comments

Anthropic recently shipped Legal Skills for Claude Desktop's Cowork, which triggered a market selloff. We ran its contract review feature against our own custom, legally engineered pipeline. The SaaS contract, playbook, and model (Opus 4.6) were all the same. The results were not: - Findings: 222 vs. 18 (12.3x delta) - Track changes applied to DOCX: 52 vs. 0 - Quantitative risk score: 93/100 vs. "CRITICAL" (a label, not a number) - Market benchmarks with percentiles: 12 vs. 0 - Comparable contracts analyzed: 4 vs. 0 - Deliverables produced: 7 vs. 1 The pipeline also produced a send-ready negotiation email, a strategy report with 10 priorities each with target/fallback/walk-away positions, 8 anticipated pushback scenarios with prepared counters, 15 talking points citing market data, and analytics with risk scoring, playbook alignment, and market percentile benchmarks for every major term. That said, Claude Desktop's output, while narrow, did correctly identify the 8 most dangerous provisions from the Provider's perspective: liability cap carve-outs, MFN pricing, escrow, uncapped indemnification. For a 5-minute investment, that's genuinely impressive. But even here, the depth gap was critical and obvious. Only the pipeline connected Section 13.3(f), which carves out insurance-covered losses from the cap, to the insurance requirements in Section 16. The practical effect: Provider's own insurance policy limits become the real liability cap, not the negotiated number in Section 13. You negotiate a $1M cap. You carry $10M in cyber insurance. Your actual exposure is $10M. A single pass reads Section 13 and Section 16 independently. Sixteen specialists cross-referencing each other catch the interaction. The same pattern repeats throughout. A 5-day outage simultaneously triggers SLA credits, DR termination, escrow source code release, and unlimited liability — a cascade across four sections. Three separate payment withholding mechanisms plus an express prohibition on service suspension effectively eliminates revenue protection. These mechanisms are not an accident. They're a coordinated conspiracy woven into the contract. And only AI specialists, working in parallel and synthesized together, are built to find it. This is our fifth controlled experiment. The pattern holds across drafting, triage, legal research, redlining, and now contract review. The model is not the product. The architecture around the model is the product. If you want to see what that looks like for your legal work: Taylor Legal Engineering, LLC. Full interactive report with methodology, pipeline architecture, and all comparison data: https://lnkd.in/eyiaXcdd #LegalTech #LegalEngineering #AI #ClaudeDesktop #ContractReview #LegalAI #LegalOps #MultiAgentAI #SaaS

25

4 Comments