AI-assisted coding has made building software faster than ever.
But speed alone does not create value.
This session explores how Product Engineering helps teams:
• Focus on user needs
• Connect engineering to outcomes
• Build intentionally, not just quickly
Catch Bethany Ann at Nebraska.Code().
Bethany Ann has a Spotlight Session at Nebraska.Code() this July.
Check out 'The Feature Nobody Used (& what it taught me about Product Engineering)' here: https://nebraskacode.amegala.com/
#ProductEngineering #FrontEnd #SoftwareCraftsmanship #Velocity #VibeCoding #AI #TechConference #WomenWhoCode #WomenInTech #WomenInSTEM #TechTalk
Stále více firem propouští produktové týmy a sází na jednu roli, která to zvládne celé sama. Product Engineer je člověk, který vymyslí produkt, implementuje ho a vyhodnotí výsledky. S ekosystémem AI agentů místo kolegů. Efektivita? Na první pohled určitě. Ale je rozdíl mezi tím dodávat víc a rychleji a skutečně být efektivní. Tenhle rozdíl firmy zatím moc neřeší.
https://zdrojak.cz/clanky/product-engineer-supermani-nebo-falesna-efektivita/
A possible solution is to shift the junior pathway away from writing lots of boilerplate code toward a trajectory like: domain -> processes -> data -> architecture -> #AI orchestration -> product.
It will not be the strong coder with a bit of domain knowledge who "wins", but the bearer of business semantics, capable of turning a domain into verifiable specifications for AI.
Anthropic identified three product bugs behind weeks of Claude Code quality complaints: a reasoning-effort downgrade, a caching bug that cleared context every turn, and a verbosity prompt that cut eval scores 3%. Shows how product-layer changes can mask as model regressions. All fixes shipped April 20, limits reset for subscribers. #AI #ProductEngineering #TechnicalDebt
Anthropic said Thursday that three product-layer changes shipped between March and April degraded Claude Code, closing out weeks of user complaints and public pushback from company staff. The company traced the drop to a March 4 reasoning-effort downgrade, a March 26 caching bug that cleared thinking blocks on every turn instead of once, and an April 16 verbosity instruction that cut coding-eval scores by 3%. All three fixes shipped by April 20 in v2.1.116.
Nathan One (@Nathanone)
AI 모델 자체보다 실제 제품화에서 더 큰 병목은 플러밍이라는 점을 지적한다. 아마존 경험을 예로 들며, 핵심 과제에서 모델이 95% 수준에 도달한 뒤에도 컨텍스트 레이어, 데이터 레이어, 권한 레이어 등 주변 시스템 작업에 예상보다 훨씬 많은 시간이 소모된다고 말한다.
@mattshumer_ @arielplk the models aren't the messy middle, the plumbing is. ive watched this at amazon, the model gets to 95% on the core task, and then the six months you thought it would take gets eaten by the context layer, the data layer, the permission layer, the "whose definition of 'customer' do
Bethany Ann, Justin Castilla, Zoe Steinkamp & Brian Gorman have Spotlight Sessions at Nebraska.Code() this July.
https://nebraskacode.amegala.com/
#Buildertrend #Postgres #ClickHouse #OpenData #Iceberg #DeltaLak #Hudi #OpenSource #ProductEngineering #FrontEnd #VibeCoding #AgenticAI #MCP #ArtificialIntelligence #TechConference #Midwest
Nebraska.Code
Anton | Access-First Auth
Zdroják
Alexei Ovsyannikov
Marcus Schuler
sayzard