My TOP 5 Agentic Engineering Bets: Anthropic, Custom Agents, & Multi-Agent UI (Not for beginners)

Detailed Summary

Introduction (0:00 - 1:07)

• IndyDevDan introduces his top five personal bets for 2026, emphasizing that planning the future helps create it.
• These bets guide his allocation of agents, time, money, and compute.
• The goal is to provide a framework for other engineers' roadmaps.
• A major announcement tying these bets together will be revealed midway through the video.

Bet 1: Custom Agents (1:08 - 3:01)

• The first bet is on custom agents, which offer a clear and untapped return on investment.
• Using custom agents in production systems immediately places engineers in the top 1% due to their ability to perform scaled, domain-specific engineering work efficiently.
• Custom agents enable repeatable problem-solving with compute by encoding solutions.
• The risk involves a steep learning curve and increased reliance on AI, but the reward is a defining competitive advantage in 2026.

Bet 2: Deprecating Old Engineering Skills (3:02 - 5:24)

• Moving forward requires letting go of outdated engineering skills, symbolized by a pencil where creation and deletion are balanced.
• The focus shifts from what an engineer can do to what they can teach their agents to do.
• Risks include breaking multi-year habits and guaranteed skill atrophy, similar to how calculators or Google Maps deprecated certain human abilities.
• Rewards include massive productivity gains and future-proofing one's career by embracing new agent-first methods.

Bet 3: Limit Breaking Agentic Coding Constraints (5:25 - 7:17)

• The two main constraints in agentic coding are planning and reviewing, which consume significant time.
• The bet is to overcome these constraints in 2026, primarily through custom agents integrated into Agentic Development Workflows (ADWs).
• This approach combines traditional engineering with new agent capabilities to unlock insane scales.
• The implementation is complex, requiring a deep understanding of the codebase, but promises a 1 to 2 order of magnitude improvement.

Bet 4: Multi-Agent UIs and Interaction Interfaces (7:18 - 9:37)

• The chat interface is identified as the laziest and most overused interface for agents.
• The focus for 2026 is on multi-agent UIs and interaction interfaces, especially voice, to increase the information rate between engineers and agents.
• This allows for wielding more compute faster, enabling systems to perform work based on an engineer's thoughts.
• Challenges include user confusion and complex UI/UX design, but the reward is another order of magnitude increase in engineering output, particularly for use-case-specific multi-agent systems.

Bet 5: Compute Maxing (9:38 - 13:37)

• Engineers are encouraged to maximize their compute usage, aiming for always-on, 24/7 agents solving problems.
• The progression involves continuously thinking about how to use more compute, moving beyond simple prompting to building complex agentic workflows.
• Risks include increased cost, architectural complexity, and potential waste, but these are mitigated by declining compute costs and improving open-source models.
• The reward is having living software that works while you sleep, achieving true agentic automation and exponential leverage.
• OpenAI's ChatGPT Pulse, which can initiate conversations, is cited as an example of systems moving towards agents prompting users.

Summary of Bets and Course Announcement (13:38 - 14:41)

• A recap of the five bets: custom agents, deprecating old skills, limit-breaking agentic coding constraints, multi-agent UIs, and compute maxing.
• These bets lead to a massive opportunity, and IndyDevDan announces "Tactical Agentic Coding" to translate these bets into daily actions.

Tactical Agentic Coding Course Details (14:42 - 24:27)

• The course, "Tactical Agentic Coding," is designed for engineers to master agentic coding tactics and scale beyond AI coding and "vibe coding."
• It's for mid to senior-level engineers who ship to production and understand that the future is agent-first.
• The course aims to transform engineers into indispensable assets by teaching them to build systems that build systems.
• It covers eight lessons and eight tactics, progressing from beginner to advanced topics like AFK agents, self-correcting systems, agentic review, and the agentic layer.
• An extended lesson package, "Agentic Horizon," offers additional content and voting rights on future lessons.
• Pricing includes early bird specials and discounts for existing "Principal AI Coding" members, with team purchasing options.
• A 30-day, no-questions-asked refund policy is offered before lesson 4.

Conclusion (24:28 - 25:35)

• IndyDevDan reiterates the importance of having a plan for the future, even if not through his course.
• He re-emphasizes his five bets and predicts a compute crunch in 2026 as engineers increasingly scale their compute usage.
• The video concludes by encouraging viewers to stay focused and keep building.

Notable Quotes

"Those who plan the future tend to create it." — IndyDevDan

"It's not about what you can do anymore. It's about what you can teach your agents to do." — IndyDevDan

"The chat interface is the laziest and most overused interface for agents." — IndyDevDan

"If you're compute maxing, you have living software that works for you while you sleep." — IndyDevDan

"At some point the conversation will flip and our agents will prompt us more than we prompt them." — IndyDevDan

"You the engineer are the bottleneck, not the models, not the tools, not the agents." — IndyDevDan

"My goal here is to turn you into an engineer they can't replace." — IndyDevDan