At the QCon New York 2023 conference, the lecture on "Optimizing Teams for Agile Survival in the Post-Agile Era" included two morning sessions that discussed how to use AI engineering and conversational software delivery to replace agile processes.
Bryan Finster gave a speech on "Agile Rehabilitation: Engineering Practices for Improved Delivery Efficiency." Finster stated that the biggest problem in software delivery is applying solutions to problems that we don't understand. He emphasized the need to think about why and what problems we are trying to solve when using agile practices. The main problem they encountered was uncertainty, which required designing solutions to provide better value.
Their approach involved deconstructing delivery using domain-driven design. They broke the system down into subdomains, defined interfaces, and studied dependencies. They then adopted a reverse Conway strategy to align the organizational structure with the desired architecture. They allocated certain parts of the system to product teams and created a small platform team.
To drive improvement, Finster outlined several practices they implemented. These practices included ensuring that system components could be deployed independently and coordinated through APIs. They also specified a target of 90% test coverage, but the results were not as expected. Finster emphasized the need to avoid meaningless tests.
To implement continuous delivery (CD), Finster created a dependency graph that described the practices outlined in the book. This graph showed which practices should be adopted to enable others. Finster mentioned that CD does involve some automation, but it is primarily about behavior.
They adopted an approach called "Minimum Viable Process." They started with no process and only added value-added content, such as refining team work, pair programming, and synchronous code reviews. They stopped doing retrospectives and instead focused on daily checks and improvements.
Finster advised against prescribing code coverage. Despite having "terrible" tests, they achieved 90% test coverage. They had some meaningless unit tests and testing lagged behind development. The main issues they encountered in testing were a lack of testing knowledge and vague requirements.
They started with Behavior-Driven Development (BDD) and defined requirements as acceptance tests, which proved effective in addressing their requirements challenges. They aimed for small work increments that could be completed within two days.
Finster mentioned that they designed their process for operations. CD exists to respond to events and safely resolve operational issues. They always used their emergency process.
One result of their approach was becoming reliable because all work was small. Another result was that their engineers regained their passion for development.
Finster shared some lessons learned:
- Use agile practices, but only use meaningful content.
- If something can be done through engineering, do it that way.
- Talent is there; people need to have the right problems to solve.
- Take CD seriously; start with continuous integration and make it work.
In their presentation "Identifying and Addressing Delivery Inefficiencies with AI," Lee Rehwinkel and Katharine Chajka delved into how AI can play a role in the software development process and be seen as a value stream agent.
Chajka presented a value stream map that revealed that the actual time spent writing code accounted for a small portion (8%) of the organization's time. She believed that even if AI minimized the time spent writing code to zero, it would hardly affect the total time spent developing new software features. This is because most of the time in the software delivery process is spent on planning, designing, testing, and releasing tasks.
Chajka demonstrated that AI can also play a significant role in managing the value stream. She proposed that AI can help identify bottlenecks in the workflow and take measures to simplify the delivery process.
Rehwinkel explained in detail how AI eliminates technical jargon in value stream management, making it more understandable throughout the organization. AI enables team members to communicate in plain language, avoiding technical terms that only those with a background in value stream management can understand.
Rehwinkel showcased a functional example of an AI value stream agent using ChatGPT. The example started with a typical query from a supervisor, such as "What should I be concerned about?" ChatGPT was able to review process metrics, identify trends, and generate analysis results. Although the analysis was detailed, it was filled with technical jargon. Rehwinkel then used a second step with ChatGPT to create a supervisor summary that transformed the information into more understandable terms. He demonstrated how ChatGPT acted as a "fact checker" to ensure that the conclusions matched the underlying data.
Rehwinkel emphasized some challenges his team faced when developing production applications using this approach. These challenges included the time-consuming response chain of ChatGPT, which could take several minutes and result in a degraded user experience. He also pointed out that these AI models are susceptible to "runaway" effects, where users can manipulate queries to change the specified role of the ChatGPT model.
(Source: https://www.infoq.com/news/2023/07/descaling-AI-development/)
Alphabet Q1 paid subscribers reached 350M, fueled by YouTube & Google One. YouTube ad revenue slightly missed at $9.88B amid Premium growth. Total Q1 revenue hit $109.9B, beating forecasts.
Anthropic releases Claude connectors for Adobe, Blender, and Ableton, enabling AI workflow automation and app control. It also sponsors the Blender Development Foundation.
Apple introduces an App Store model splitting 12-month plans into monthly installments. On iOS 26.4+, it lowers upfront costs without reducing total price, adding billing tracking.
Google launches Aadhaar digital IDs in Wallet for India, enabling secure QR/NFC verification. A phased partner rollout targets key sectors, while passport-based credentials expand globally.
YouTube tests "Ask YouTube," an AI tool blending text and video for complex queries. Initially for US Premium users, it supports interactive follow-ups, expanding Google's AI search.
Google grants US DoD classified AI access, joining OpenAI and xAI in defense deals amid Anthropic's dispute. Contractual safeguards against surveillance and lethal weapons remain unenforceable.
Mistral AI launches Workflows, a Temporal-powered orchestration platform for secure, Python-driven enterprise AI automation featuring human-in-the-loop approvals.