Leadership at Scale; ChatGPT is not AI; Building Winning Teams; 8 AI Uses for EMs
Issue #58 Bytes
🌱 Dive into Learning-Rich Sundays with groCTO ⤵️
Leadership at Scale: From Chaos to Empowerment
Ever feel like the bigger your team gets, the more you're just… managing chaos instead of leading? It's a common trap: as teams expand, leaders often get bogged down in complexity, becoming the very bottleneck they're trying to avoid.
But what if I told you there's a different way? After years of guiding teams through the wild ride of software delivery, I've discovered that true leadership isn't about pulling all the strings or dictating every move. It's about something far more powerful, and honestly, a lot more satisfying. Dive into the full article and unlock the secrets to building truly independent, high-performing teams👇
Editorial note: We’re aware of the METR AI Impact Study circling around social media regarding AI slowing engineers down. As it is laden with bias we’ll cover it at a later point once the dust settles. In the meantime you can read the DORA group discussion on it here and give you instead a rundown about ChatGPT in today’s featured article. Thank you for being patient and enjoy the summer!
Article of the Week ⭐
“[…] ChatGPT predicts patterns, not facts. When it encounters a gap in its training data, it fills it with whatever seems statistically likely. The model lacks a mechanism to indicate "I don't know"; it simply continues to predict the next plausible token.“
ChatGPT is not AI
Dr. Milan Milanović breaks down what ChatGPT is made of
People conflate ChatGPT with artificial intelligence itself. Dr. Milan sets the record straight: ChatGPT is one expression of AI-like models, not the whole discipline.
What You Should Know
AI is a broad field: It includes everything from robotics and computer vision to expert systems and ML (which provide the theory under the hood of chatbots).
ChatGPT sits at the end of a deep tech hierarchy: AI → Neural Nets → Deep Learning → Transformers → LLMs → GPT → ChatGPT, the application service.
It’s just one tool among many: A conversational UI layered over GPT-4, tuned with reinforcement learning and optimized for fluency, although many would like to believe it to be close to general intelligence, which it is not (yet?).
How ChatGPT Works
Training pipeline: ChatGPT starts with unsupervised learning, predicting the next token across billions of examples. This is followed by supervised fine-tuning, then Reinforcement Learning from Human Feedback (RLHF) to shape tone, safety, and relevance.
Transformer architecture: It uses a decoder-only model with multi-head self-attention, allowing it to weigh different words in context, rather than processing language one token at a time like older RNNs.
Token-by-token generation: It doesn’t know the destination when it starts a sentence. Each word is chosen based on the statistical likelihood of following the prior words. The result sounds coherent, but it’s reactive, not reflective.
No plan, no reasoning: There’s no intent behind the response. It doesn’t remember three paragraphs ago unless the context window includes it, and even then, it lacks a mental model of consistency.
Hallucinations: Because the model generates rather than retrieves, it will sometimes invent convincing but false details. Especially when pushed for specifics. This makes validation essential in high-stakes use.
Why It Matters
If you treat ChatGPT like a magic oracle, you’re setting yourself up for trouble. Understand its strengths (speed, fluency, summarization) and weaknesses (hallucination, lack of reasoning). Knowing what’s under the hood helps you use it wisely and explain it accurately to your team or execs.
Other highlights 👇
How to Build Winning Engineering Teams
Engineering leaders get specific about what matters most when hiring. In his Executive Engineering interview roundup Yassine Kachchani asked five seasoned tech leaders to weigh in on the traits that make or break a team. Great teams are made of people who ship, learn, and lift others around them.
Spoiler: It’s not about rockstars, it’s about reliability, self-awareness, and people who take initiative without being told.
Key Qualities to Hire For
1. Team spirit over ego.
A high-IQ engineer who can’t collaborate is a liability. Mourad Zerroug (PlanHub) emphasized team-first attitudes, initiative, and low overhead: “Engineers who drop tasks or need frequent reminders show no passion to play an active role.”
2. Ownership mindset.
Whether it’s Pogacsas’ emphasis on self-awareness or Stanek’s need for engineers who act like product owners, there’s one common theme: engineers who care about outcomes, not checklists.
3. Learning and adaptability.
Top engineers are never done learning. The best ones actively seek growth, evolve with the stack, and transfer energy to others.
Hiring is a Relationship
Emese Pogacsas draws a smart comparison: hiring is more like dating than shopping. “You need multiple encounters to really know if there’s a match.” Multi-step processes help both sides figure out compatibility, values, and communication style far beyond what a resume can show.
There’s No Perfect Team
Kevin Goldsmith (DistroKid) notes that there’s no one-size-fits-all team. It’s about finding the right mix of skills and personalities for the problem space you're in now. Balance matters more than brilliance.
Remote Work Demands Builders
Mirek Stanek (Papaya Global) stressed one thing: remote teams only work when engineers go beyond the ticket. You need people who simplify chaos, collaborate cross-functionally to reduce complexity, and focus on real customer outcomes without needing a map.
Why Coding Challenges Still Matter
Emily Nakashima (Honeycomb.io) defends their coding tests: not as perfection filters, but pressure checks. In early-stage environments, progress matters more than polish. How candidates handle constraints often mirrors how they’ll handle your roadmap.
8 Daily Ways I use AI as an Engineering Manager to Level Up my Work
A hands-on rundown from Caleb Mellas on how AI saves time, improves quality, and fits into a real daily workflow. Engineering managers juggle context like standups, incidents, strategy docs, onboarding, and that never-ending block of back-to-backs. Caleb walks through exactly how he uses AI to handle the chaos without burning out.
Everyday Wins with LLMs
Here’s where AI has delivered real leverage for Caleb in his day-to-day:
→ SQL Reporting & Dashboards
Generating cross-table queries and tweaking Snowflake dashboards is faster with Claude. Caleb uses it to draft queries, explain tricky functions like DATE_TRUNC, and format output for stakeholders. Still requires human review, but now it’s iterative, not from scratch.
→ Excel/Sheets Analysis
Need to clean, merge, or dedupe 120K rows across messy phone columns? AI speeds up formula creation, from counting non-nulls to isolating mismatches across datasets. Gemini helps generate, adapt, and explain complex formulas faster than writing custom scripts.
→ Debugging State Leaks
Tests failing without clear reasons? Instead of hours of solo investigation, Caleb used .skip + .only to isolate the issue, then had Gemini confirm the cause to be shared mutable state between tests. AI helped validate his instinct and offer a precise fix.
→ Refactoring Legacy Code
Refactoring older components to match modern patterns is where AI shines. Gemini got Caleb 80–90% of the way through migrating Vue components, controllers, and tests—fast enough to fit between meetings and avoid another backlog item.
→ Writing Tests
No time for TDD? AI drafts Jest test suites from code examples and validates assumptions on the fly. With some nudging on TypeScript types and interface definitions, Caleb adds coverage without sacrificing velocity.
→ Mermaid Diagrams
Production issue? Complex flows across teams? AI-generated diagrams help document and communicate what broke and where. Mermaid + Claude = fast visual clarity for PMs, devs, and stakeholders.
→ Incident TLDRs
Too deep in Slack threads to summarize an incident? AI extracts key facts—what happened, when, impact, what’s been tried, and what’s next. Caleb uses this to quickly sync teams and stakeholders during high-pressure moments.
→ Docs That Don’t Suck
LLMs aren’t great doc writers by default. But with code context, preferred examples, and a back-and-forth edit mindset, they become solid co-writers. Caleb shares prompts and practices that cut writing time without sacrificing clarity or accuracy.
Find Yourself 🌻
That’s it for Today!
Whether you’re innovating on new projects, staying ahead of tech trends, or taking a strategic pause to recharge, may your day be as impactful and inspiring as your leadership.
See you next week(end), Ciao 👋
Credits 🙏
Curators - Diligently curated by our community members Denis & Kovid
Featured Authors - Dr. Milan Milanović, Yassine Kachchani, Caleb Mellas
Sponsors - This newsletter is sponsored by Typo AI - Ship reliable software faster.
1) Subscribe — If you aren’t already, consider becoming a groCTO subscriber.
2) Share — Spread the word amongst fellow Engineering Leaders and CTOs! Your referral empowers & builds our groCTO community.