Hi Everyone, We’re back with the sixth edition of Expert Talks for the year.
Please join us on 24th Oct 2026, from 10:30 AM to 1:00 PM @ Equal Experts (11th Floor), IndiQube Alpine, Ekkatuthangal, Chennai.
Agenda : 1. Welcome & Intros (10 mins) 2. Talk 1: Signal Over Slop: The Case for Context Engineering 3. Tea Break (15 mins) 4. Talk 2: From Boomerang to Avatar State: Bending Claude to Your Codebase
Talk 1: Signal Over Slop: The Case for Context Engineering by Niranjani
When a major AI lab's model started calling itself a goblin mid-conversation, it made headlines as a funny glitch. It's actually a symptom: models act on what's literally in their context, not what we intended them to understand. As AI agents move from pair-programmer to autonomous task-runner, that gap between intent and context widens invisibly - until "context rot" degrades an entire codebase.
Context engineering isn't a prompt-writing trick; it's the architectural discipline that keeps a codebase's relationship with AI tools maintainable over time. Without deliberate guardrails, developers end up in unscoped, endless chat sessions that quietly compound system debt and inject unreviewed "slop" into production repositories.
This talk skips beginner prompt tips for concrete, load-bearing patterns: citation discipline, output constraints, progressive disclosure, durable state, and checkpointing.
Outline
1. Introduction Models act on what's actually in context, not what we intended 2. Demystifying the terms and symptoms a) Emergent training artifacts, context rot, and AI slop as the same failure shape at three layers: training, context window, and codebase b) Slop as the downstream cost: how unmanaged context quietly degrades codebase health over months 3. The Context Engineering Playbook 5 Patterns for keeping agent context under control 4. Demo: Watching Context Accumulate a) A real-time look at exactly where things go sideways. b) Identifying where noise creeps in and what would have helped 5. Challenges, What's Next, & Q&A Where these patterns get harder and more necessary - once agents start running multi-step tasks without a human watching every turn.
Talk 2: From Boomerang to Avatar State: Bending Claude to Your Codebase by Raja Syed
Turns out the discipline behind using AI effectively was hiding in the world's favorite animated series all along. This practical session offers a manual for working with Claude across the software development lifecycle (SDLC), structured around four iconic elements from Avatar: The Last Airbender.
Outline
Setting the Scene 1. Quick refresher on the Avatar: The Last - Airbender and how it relates to this session. - This talk is about using AI to build, not AI inside what you build - Capability is solved, aim is not 2. Sokka's Boomerang - When not to use AI - Tasks with one correct answer belong to deterministic tools, not the model 3. Katara's Waterbending - Claude takes the shape of whatever's in the room with it - Give it your codebase's shape once, in writing, and it stops guessing 4. Toph's Teaching - Skills turn one person's know-how into written instructions - Any new session can use them: the knowledge isn't locked to one person 5. The Avatar State - Full power with no aim is exactly what an unsupervised agent looks like - Permission modes: how much freedom you give it before it acts at all - Hooks: the one rule that holds even when the model disagrees.
About the Speakers
Niranjani is a lead software engineer with 10 years of experience helping teams modernize legacy systems and build scalable distributed applications."She's worked across public health, travel, and retail domains - mostly JVM stacks, with the occasional frontend detour. A fan of evolutionary architecture and clean software design, Niranjani enjoys organizing community meetups and sharing what she learns through talks and team lunch-and-learns.
Raja Syed is working as a senior software engineer in Equal Experts. Throughout his career, he has worked across the full stack from frontend and backend development to microservices, cloud platforms, and data engineering. Being a generalist has given him a deep perspective on system interconnectivity and technical tradeoffs. Having watched the shift from Stack Overflow to AI-native development tools, He is passionate about helping engineers navigate how AI is fundamentally reshaping software creation.
