James Asbury
@jbury.bsky.social
Software engineer @ Microsoft
Views expressed are my own.
San Francisco, California 🌁
Views expressed are my own.
San Francisco, California 🌁
Reposted by James Asbury
I run all of my DNS queries through a local LLM because it’s more deterministic
October 21, 2025 at 5:37 AM
I run all of my DNS queries through a local LLM because it’s more deterministic
Reposted by James Asbury
it’s really interesting to see how the popular conception of AI is exclusively the chat products
There is no conception of things like an LLM acting as an engine for a different sort of product
it would be like dismissing the internet on the basis of not taking AOL chat seriously
There is no conception of things like an LLM acting as an engine for a different sort of product
it would be like dismissing the internet on the basis of not taking AOL chat seriously
October 7, 2025 at 1:27 PM
it’s really interesting to see how the popular conception of AI is exclusively the chat products
There is no conception of things like an LLM acting as an engine for a different sort of product
it would be like dismissing the internet on the basis of not taking AOL chat seriously
There is no conception of things like an LLM acting as an engine for a different sort of product
it would be like dismissing the internet on the basis of not taking AOL chat seriously
Reposted by James Asbury
very relatable, honestly
August 30, 2025 at 7:27 PM
very relatable, honestly
I see a lot of rhetoric calling AI a “fad” that will fade away like NFTs a few years back.
This is missing a huge point. The next generation of college graduates find AI tools to be second nature. When you become comfortable with a technology that makes you feel more productive, you can’t go back.
This is missing a huge point. The next generation of college graduates find AI tools to be second nature. When you become comfortable with a technology that makes you feel more productive, you can’t go back.
August 7, 2025 at 4:17 PM
I see a lot of rhetoric calling AI a “fad” that will fade away like NFTs a few years back.
This is missing a huge point. The next generation of college graduates find AI tools to be second nature. When you become comfortable with a technology that makes you feel more productive, you can’t go back.
This is missing a huge point. The next generation of college graduates find AI tools to be second nature. When you become comfortable with a technology that makes you feel more productive, you can’t go back.
Somewhere here is a reality that hits close to home for today’s developers (for better or worse).
Going forward we will see an abundance of college graduates who find AI tools second nature. Developers neglecting these skills will be seen as slower/falling behind
en.m.wikipedia.org/wiki/Who_Mov...
Going forward we will see an abundance of college graduates who find AI tools second nature. Developers neglecting these skills will be seen as slower/falling behind
en.m.wikipedia.org/wiki/Who_Mov...
August 6, 2025 at 3:45 PM
Somewhere here is a reality that hits close to home for today’s developers (for better or worse).
Going forward we will see an abundance of college graduates who find AI tools second nature. Developers neglecting these skills will be seen as slower/falling behind
en.m.wikipedia.org/wiki/Who_Mov...
Going forward we will see an abundance of college graduates who find AI tools second nature. Developers neglecting these skills will be seen as slower/falling behind
en.m.wikipedia.org/wiki/Who_Mov...
Reposted by James Asbury
Dear AI coding helper, Sometimes a tab is just a tab and not a request to rewrite the line of code.
June 16, 2025 at 5:54 PM
Dear AI coding helper, Sometimes a tab is just a tab and not a request to rewrite the line of code.
Respectfully disagree. There is no “one size fits all” approach to testing. Start with a thoughtful analysis of where regressions, bugs, and edge cases are likely to arise. Then pick the best way to guard against them for your project.
Write flat tests. No "describe", no scenario nesting, no beforeEach/afterEach hooks (as a rule). Each test case is self-contained, listing all the setup it needs. Use Symbol.dispose for reliable cleanup. Use granular test cases.
☝️ These things will make your tests shine.
☝️ These things will make your tests shine.
June 15, 2025 at 12:24 AM
Respectfully disagree. There is no “one size fits all” approach to testing. Start with a thoughtful analysis of where regressions, bugs, and edge cases are likely to arise. Then pick the best way to guard against them for your project.
Reposted by James Asbury
