AI
The State of AI Software Development: Data from 400+ Orgs
- Deployment frequency up about 20% over the last year
- Perceived speed is up about 4%
- Change failure rate is volatile (several improved, several got worse); the amplitude is bigger over the last year
- Code maintainability is up about 4%, but change confidence is down about 6%. It’s easier for me to make changes with AI, but I trust the output less.
- PR size increased about 63%
- Incremental delivery sentiment has gone down about 10%
- Junior engineers are using AI the most; senior engineers use fewer tokens
- The amount of time saved by using AI is about the same regardless of seniority
- Time saved by using AI is slightly higher in smaller companies
- For DX, core metrics = measures whether it’s working, AI metrics = measure what’s happening (utilization, impact, and cost)
- Good developer experience (maturity of engineering practices in your code) leads to better performing agents
- SDLC/PDLC integration… “Code generation was never the bottleneck in the first place.” Companies are not getting massive multipliers like people were promised.
- Biggest non-AI factors that eat up time are meeting heavy-days and interruptions
- Doomers
- Extropians, one of whom (Eliezer Yudkowsky) said that super-intelligent AI is inevitable but having a positive outcome for humanity was not guaranteed. This group is the rationalists.
- Effective altruism (EA) is another group that’s recently focused on preventing the extinction of humanity. (This group focuses on other things than just AI.)
- The EA and rationalists are almost indistinguishable in the San Francisco Bay area.
- Anthropic & OpenAI. OpenAI was started in large part to fulfill the idea that we need to align AI to prevent extinction because they didn’t trust Google (Deep Mind) to do it the right way. Anthropic splintered off OpenAI once it went for-profit.
- There are EA/rationalist folks who are very upset at Anthropic and OpenAI for moving too fast.
- This group is well funded and is lobbying Congress to shape the kind of regulation they want.
- Accelerationists
- Believe the Doomers are weird and don’t see why we’d slow the benefits of AI combined with the economic race with China.
- Venture capitalists (e.g., Marc Andreesson)
- Other AI firms (e.g., Mark Zuckerberg, Jensen Huang)
- US White House
- The motive here is money and power (don’t ban chip exports, don’t impede startup growth).
- Centrists
- Concerned about AI and are in favor of regulation (like any other unsafe tech), but aren’t interested in the doomer philosophies.
- AI ethicists — focused on proximate harms (biased data, surveillance of minorities). See also TESCREAL.
- Normal technology computer scientists — focused on understanding what the technology does
- Cybersecurity experts — they don’t care for the superintelligence fairy tale, and know that the hacking tools are very dangerous.
- This group is concerned that Doomer-inspired philosophy is driving how regulation will look.
Software development
Azure Application Insights with .NET Core Web API
This brief tutorial shows how to set up Application Insights and connect it to a .NET Core Web API to establish observability.