Peppe's Working Notes
Search
Ctrl K
Peppe's Working Notes
Search
Ctrl K
MAIN NOTES
HUB NOTES
$Building software products with AI
$Designing Agentic Workflows
$Evolution of roles in software engineering
$Organizational Dynamics & Structure
$Software Complexity & Practice Evolution
"LLM as a judge" grading with a rubric gives more consistent results when evaluating Reflection (Agentic pattern)
A new skill can be learned starting from a problem to solve
A team is not a group of people
Agentic AI can be less or more autonomous
Agentic AI works better for well know, linear processes
Agentic design patterns
Agentic workflows are only as good as their weakest component + error analysis
Agents can be evaluated objectively or subjectively
Agile practices only make sense when iteration is cheap
Autonomy without structure consistently produces informal hierarchy
Brainstorming only works with preconceived knowledge
Coding with LLMs makes me think more
Context compression for better code generation
Continuous delivery of software often translates into "let's release whatever works without properly architecting the software"
Distributed decision-making in a traditional hierarchy is usually dead on arrival because of legacy power structures
Hiring product engineers like product managers
Methods, methodologies and mindset
Peppe's Working Notes
Planning (Agentic Pattern) with code improves performance
Product Engineers are a rejection of fragmentation
Product engineers with strong industry knowledge can communicate better in that domain
Product engineers with strong industry knowledge do better problem solving in that domain
Product engineers with strong industry knowledge make better technology choices
Product outcomes differ from business outcomes because they can be controlled by the product team
Pull Requests are just a tool
Reflection (Agentic Pattern) can use external feedback from tools to improve output
Reflection (Agentic Pattern) consistently outperforms direct generation on a variety of tasks
Slow feedback loops and expensive computing forced optimization for machines over people
Software development is complex, not complicated
Software engineers should be involved in product decisions
Software engineers won't need to write and read code only when agents will embed good engineering practices in their DNA
Software systems integrating with AI still benefits from good software architecture
The atomic unit of what happens in the workplace is not the job, but the task
The correct response to complexity is probe, sense, respond
The currency of the future will be thought
The increase in uncertainty drove each era of software development practices
The PM role fills the vacuum left by a lack of clear organisational systems.
The PM role is necessary at scale as a coordination role
There are no best practices in complex systems
There's not difference in maintainability in AI vs Human generated code
We should invest in AX (Agents Experience)
Years of experience are not a good metric to assess someone's ability to do the job
REFERENCE NOTES
Marc Andreessen, The real AI boom hasn’t even started yet
Select a result to preview
Reflection (Agentic Pattern) consistently outperforms direct generation on a variety of tasks