writing
Writing
Essays on engineering leadership, mentorship, and how AI changes how teams collaborate and ship.
$ ls260 items
- Engineering leadership
- When retrieval lies politely: what fails first
- Multi-agent theater vs useful specialization: naming the owner
- The meeting you should replace with a doc: evidence over enthusiasm
- When a vendor becomes your process: a shorter feedback loop
- Designing for fewer retries, not prettier dashboards: evidence over enthusiasm
- When leadership means clearing drag, not adding process: the operable version
- Hiring for ownership, not theatrical confidence: a shorter feedback loop
- Sustainability reviews in architecture forums: evidence over enthusiasm
- Evaluating AI assistance without vanity metrics: naming the owner
- Connectors are the hard part of retrieval: naming the owner
- Teaching judgment without becoming a bottleneck: what fails first
- Efficiency as reliability practice: a shorter feedback loop
- Growing seniors who can grow others: evidence over enthusiasm
- Mentoring in public without performing: a shorter feedback loop
- What Asim Hussain’s green software work asks of us: what fails first
- Trust compounds slower than roadmaps: what fails first
- Owning outcomes when agents ship changes
- When a vendor becomes your process: evidence over enthusiasm
- Hand-offs that preserve context: evidence over enthusiasm
- Career conversations that change next week’s work: evidence over enthusiasm
- Collaboration tax and how to cut it: what fails first
- RAG that fails gracefully when context is thin
- Hiring for ownership, not theatrical confidence: evidence over enthusiasm
- Exit strategies for bought AI tooling: keeping it teachable
- Connectors are the hard part of retrieval
- When leadership means clearing drag, not adding process: what fails first
- Pairing as coaching, not surveillance: naming the owner
- Multi-agent theater vs useful specialization
- When retrieval lies politely: naming the owner
- Build the harness, buy the commodity: evidence over enthusiasm
- Grounding answers in systems teams actually run
- Collaboration tax and how to cut it: naming the owner
- Growing seniors who can grow others: the operable version
- Evaluating AI assistance without vanity metrics
- Integration debt after the purchase: naming the owner
- Orchestration without a black-box conductor: naming the owner
- Teaching judgment without becoming a bottleneck: naming the owner
- Reorganizing when the product shape changes: evidence over enthusiasm
- Permissions in RAG: the boring requirement
- Idempotency for agentic side effects: what fails first
- Teaching judgment without becoming a bottleneck
- Eval loops for retrieval quality
- Feedback that lands after incidents
- The new cost of custom: maintenance, not typing — naming the owner
- AI literacy as a coaching problem: naming the owner
- When an agent should stop and ask
- When product and platform disagree productively: evidence over enthusiasm
- When retrieval lies politely
- Hiring for ownership, not theatrical confidence: the operable version
- Cross-team work without a shared roadmap: keeping it teachable
- Tool use as a privilege, not a default
- Ceremony is optional; continuity is not: naming the owner
- Working agreements that age well: evidence over enthusiasm
- Reorganizing when the product shape changes: the operable version
- Exit strategies for bought AI tooling: a shorter feedback loop
- Sustainability reviews in architecture forums: the operable version
- Coaching through ambiguity without stealing the work: naming the owner
- Career conversations that change next week’s work: the operable version
- Integration debt after the purchase
- Judgment under load: what managers actually practice — the operable version
- Measuring compute waste in product teams
- Efficiency as reliability practice: evidence over enthusiasm
- Exit strategies for bought AI tooling: evidence over enthusiasm
- RAG pipelines as team infrastructure: evidence over enthusiasm
- Idempotency for agentic side effects: naming the owner
- Carbon-aware scheduling without the slogans: evidence over enthusiasm
- Cross-team work without a shared roadmap: a shorter feedback loop
- Build less, own more of the outcome: the operable version
- Growing seniors who can grow others: what fails first
- The meeting you should replace with a doc: the operable version
- Right-sizing infrastructure as leadership work: the operable version
- AI literacy as a coaching problem
- Interfaces between teams beat shared Slack channels: evidence over enthusiasm
- Cross-team work without a shared roadmap: evidence over enthusiasm
- Buy vs build when AI compresses build time: what fails first
- Event-driven agentic workflows teams can operate
- From demo RAG to operable RAG: what fails first
- When to sponsor, when to step back: the operable version
- Reorganizing when the product shape changes: what fails first
- Designing hand-offs between agents and humans: the operable version
- Hand-offs that preserve context: the operable version
- Carbon-aware scheduling without the slogans: the operable version
- RAG pipelines as team infrastructure: the operable version
- Mentorship that survives calendar chaos: what fails first
- Shared ownership across org boundaries: what fails first
- Leaders who stay close to the craft: the operable version
- Alignment is a protocol, not a meeting: keeping it teachable
- When a vendor becomes your process: the operable version
- Coaching through ambiguity without stealing the work
- The energy cost of flaky CI: naming the owner
- Designing hand-offs between agents and humans: what fails first
- Coaching juniors through AI-assisted workflows: what fails first
- RAG pipelines as team infrastructure: what fails first
- Shipping with models without losing craft: naming the owner
- Prompting is not process — harnesses are
- Trust compounds slower than roadmaps: naming the owner
- Agent harnesses: logs, budgets, and kill switches — naming the owner
- AI in the workflow, not on the slide deck
- Designing for fewer retries, not prettier dashboards: the operable version
- Orchestration without a black-box conductor
- Hiring for ownership, not theatrical confidence: what fails first
- Exit strategies for bought AI tooling: the operable version
- Working agreements that age well: the operable version
- Keeping humans accountable when models draft code: naming the owner
- Sustainability reviews in architecture forums: what fails first
- Mentoring in public without performing: evidence over enthusiasm
- Shipping with models without losing craft
- When a vendor becomes your process: what fails first
- Alignment is a protocol, not a meeting: a shorter feedback loop
- Build less, own more of the outcome: what fails first
- Idempotency for agentic side effects
- Designing hand-offs between agents and humans: naming the owner
- Growing seniors who can grow others: naming the owner
- Build the harness, buy the commodity: the operable version
- AI changes tempo; leadership still owns outcomes
- RAG pipelines as team infrastructure: naming the owner
- Who owns the knowledge graph behind the bot?: naming the owner
- When to sponsor, when to step back: what fails first
- Leaders who stay close to the craft: what fails first
- Teaching teams to distrust fluent wrongness
- Keeping humans accountable when models draft code
- Cross-team work without a shared roadmap: the operable version
- Judgment under load: what managers actually practice — what fails first
- When to sponsor, when to step back: naming the owner
- The psychology of asking for help in engineering: naming the owner
- From demo RAG to operable RAG: naming the owner
- Efficiency as reliability practice: the operable version
- From demo RAG to operable RAG
- Who owns the knowledge graph behind the bot?
- Designing hand-offs between agents and humans
- What Asim Hussain’s green software work asks of us: naming the owner
- Efficiency as reliability practice: what fails first
- When to sponsor, when to step back
- The new cost of custom: maintenance, not typing
- Mentorship that survives calendar chaos: naming the owner
- Agent harnesses: logs, budgets, and kill switches
- Mentorship that survives calendar chaos
- Interfaces between teams beat shared Slack channels: the operable version
- Alignment is a protocol, not a meeting: evidence over enthusiasm
- Exit strategies for bought AI tooling: what fails first
- RAG pipelines as team infrastructure
- Alignment is a protocol, not a meeting: the operable version
- Build less, own more of the outcome: naming the owner
- Hand-offs that preserve context: what fails first
- Interfaces between teams beat shared Slack channels: what fails first
- Build the harness, buy the commodity: what fails first
- Designing for fewer retries, not prettier dashboards: what fails first
- Coaching juniors through AI-assisted workflows: naming the owner
- Internal open source across product teams: naming the owner
- Interfaces between teams beat shared Slack channels: naming the owner
- Contributing without cosplay: what fails first
- Contributing without cosplay: naming the owner
- Designing for fewer retries, not prettier dashboards: naming the owner
- The energy cost of flaky CI
- Architecture decisions you can reverse: naming the owner
- Open source as mentorship at scale
- Architecture decisions you can reverse
- Open source governance for small teams: naming the owner
- Cross-team work without a shared roadmap: what fails first
- Ceremony is optional; continuity is not
- When leadership means clearing drag, not adding process: naming the owner
- Maintainer burden and team capacity: what fails first
- Carbon-aware scheduling without the slogans: what fails first
- The meeting you should replace with a doc: what fails first
- Platforms you buy still need owners: naming the owner
- Collaboration tax and how to cut it
- Upstream first as a collaboration habit
- Carbon-aware scheduling without the slogans: naming the owner
- When a vendor becomes your process: naming the owner
- Coaching juniors through AI-assisted workflows
- Leaders who stay close to the craft: naming the owner
- Documentation that earns its keep: naming the owner
- Mentoring in public without performing: the operable version
- Code review as teaching, not gatekeeping: what fails first
- Testing strategy when speed matters: naming the owner
- API design for humans who inherit it: naming the owner
- When product and platform disagree productively: the operable version
- Mentoring in public without performing: what fails first
- Judgment under load: what managers actually practice — naming the owner
- Sustainability reviews in architecture forums: naming the owner
- Licenses, liability, and leadership judgment: naming the owner
- Cross-team work without a shared roadmap: naming the owner
- Contributing without cosplay
- Alignment is a protocol, not a meeting: what fails first
- When product and platform disagree productively: what fails first
- Ownership is a design problem, not a pep talk
- Incident response that improves the system
- Right-sizing infrastructure as leadership work: what fails first
- Shared ownership across org boundaries: naming the owner
- Conflict as a design input: what fails first
- Career conversations that change next week’s work: what fails first
- Working agreements that age well: what fails first
- Platforms you buy still need owners
- The quiet work of keeping delivery honest: naming the owner
- Internal open source across product teams
- Shared ownership across org boundaries
- Working agreements that age well: naming the owner
- Small platforms that stay operable
- Reorganizing when the product shape changes: naming the owner
- Technical debt with a repayment schedule
- Migrations that respect team capacity: naming the owner
- What Asim Hussain’s green software work asks of us
- Hiring for ownership, not theatrical confidence: naming the owner
- When product and platform disagree productively: naming the owner
- Designing for fewer retries, not prettier dashboards
- Buy vs build when AI compresses build time: naming the owner
- Testing strategy when speed matters
- Hand-offs that preserve context: naming the owner
- Pairing as coaching, not surveillance
- Documentation that earns its keep
- Licenses, liability, and leadership judgment
- Efficiency as reliability practice: naming the owner
- Build less, own more of the outcome
- Conflict as a design input: naming the owner
- Code review as teaching, not gatekeeping: naming the owner
- Alignment is a protocol, not a meeting: naming the owner
- Growing seniors who can grow others
- Career conversations that change next week’s work: naming the owner
- Hand-offs that preserve context
- Cross-team work without a shared roadmap
- Buy vs build when AI compresses build time
- When a vendor becomes your process
- Efficiency as reliability practice
- Open source governance for small teams
- The psychology of asking for help in engineering
- Judgment under load: what managers actually practice
- The meeting you should replace with a doc: naming the owner
- Interfaces between teams beat shared Slack channels
- Right-sizing infrastructure as leadership work: naming the owner
- Hiring for ownership, not theatrical confidence
- Docs PRs as onboarding practice
- When product and platform disagree productively
- Mentoring in public without performing: naming the owner
- Migrations that respect team capacity
- Build the harness, buy the commodity: naming the owner
- Carbon-aware scheduling without the slogans
- The quiet work of keeping delivery honest
- When leadership means clearing drag, not adding process
- Mentoring in public without performing
- Maintainer burden and team capacity: naming the owner
- Borrowing patterns, not brand names
- Maintainer burden and team capacity
- Leaders who stay close to the craft
- Exit strategies for bought AI tooling: naming the owner
- When to fork and when to fund
- Alignment is a protocol, not a meeting
- The meeting you should replace with a doc
- Working agreements that age well
- Right-sizing infrastructure as leadership work
- Trust compounds slower than roadmaps
- Exit strategies for bought AI tooling
- Reorganizing when the product shape changes
- API design for humans who inherit it
- Sustainability reviews in architecture forums
- Career conversations that change next week’s work
- Build the harness, buy the commodity
- Code review as teaching, not gatekeeping
- Conflict as a design input
- Make-or-buy with agents in the mix