there were plenty of labor saving automations in writing code before LLMs happened; we do codegen for a billion things, and I’m yet to see a cogent argument for why LLMs are a categorical change rather than a technology that’s being deployed by bad political actors with huge environmental harms
November 9, 2026 at 11:38 PM
there were plenty of labor saving automations in writing code before LLMs happened; we do codegen for a billion things, and I’m yet to see a cogent argument for why LLMs are a categorical change rather than a technology that’s being deployed by bad political actors with huge environmental harms
i think the profession model for software engineering in a world where production grade software is created by llm codegen for both development and hardening might look closer to the daily life and also pay grade of civil and mechanical engineers Paperwork & Meetings Wise but with less liability
Like many cases where AI exacerbates an existing condition, one thing going on here is the tech industry already massively failed as a business culture to engage in minimally effective industrial and organizational knowledge management with widepsread tacit knowledge loss driven by employee turnover
November 9, 2026 at 2:50 AM
i think the profession model for software engineering in a world where production grade software is created by llm codegen for both development and hardening might look closer to the daily life and also pay grade of civil and mechanical engineers Paperwork & Meetings Wise but with less liability
ai codegen is bad for a whole host of reasons but its especially baffling in fan mods/game spaces because just, like, what are we doing here if you don't have enough passion for what you do to do it yourself?
November 7, 2026 at 11:58 PM
ai codegen is bad for a whole host of reasons but its especially baffling in fan mods/game spaces because just, like, what are we doing here if you don't have enough passion for what you do to do it yourself?
📦 tinvar/codegen v0.4.15
Generates TypeScript from Tinvar Data objects, queries and routes.
🔗 https://github.com/tinvar/codegen
Generates TypeScript from Tinvar Data objects, queries and routes.
🔗 https://github.com/tinvar/codegen
November 8, 2026 at 10:59 PM
📦 tinvar/codegen v0.4.15
Generates TypeScript from Tinvar Data objects, queries and routes.
🔗 https://github.com/tinvar/codegen
Generates TypeScript from Tinvar Data objects, queries and routes.
🔗 https://github.com/tinvar/codegen
A TypeScript upgrade touches more than package.json. I'd check lint, codegen and build-plugin compatibility before switching. Compiler speed is only one part of readiness.
November 10, 2026 at 3:30 PM
A TypeScript upgrade touches more than package.json. I'd check lint, codegen and build-plugin compatibility before switching. Compiler speed is only one part of readiness.
trend: Huge jump in TPU deliveries
likely the compute is going to
1. Anthropic
2. Google
3. AI dev process — RSI, codegen, etc.
likely the compute is going to
1. Anthropic
2. Google
3. AI dev process — RSI, codegen, etc.
Broadcom TPU compute shipments doubled from fiscal Q2 to Q3 2026 (May to early August), reaching the equivalent of nearly 2 million Nvidia H100 GPUs, according to our estimates.
This is a sharp increase from the previous three quarters, when shipments grew between 2% and 22% per quarter.
This is a sharp increase from the previous three quarters, when shipments grew between 2% and 22% per quarter.
November 5, 2026 at 9:15 PM
trend: Huge jump in TPU deliveries
likely the compute is going to
1. Anthropic
2. Google
3. AI dev process — RSI, codegen, etc.
likely the compute is going to
1. Anthropic
2. Google
3. AI dev process — RSI, codegen, etc.
#TIL Playwright hat - neben dem `--ui`-Modus - auch einen codegen Recorder playwright.dev/docs/codegen #playwright #testing (Weiß gar nicht, ob cypress den codegen-Modus auch hat 🤔 )
November 6, 2026 at 12:24 PM
#TIL Playwright hat - neben dem `--ui`-Modus - auch einen codegen Recorder playwright.dev/docs/codegen #playwright #testing (Weiß gar nicht, ob cypress den codegen-Modus auch hat 🤔 )
A faster compiler doesn't mean your toolchain is ready. Before a TypeScript 7 migration, check the tools that import typescript: lint, codegen and build plugins.
November 10, 2026 at 2:30 AM
A faster compiler doesn't mean your toolchain is ready. Before a TypeScript 7 migration, check the tools that import typescript: lint, codegen and build plugins.
Everyone's favorite winning here 🫠🫠
(also lol @ the answer "Rust" 🙃)
(also lol @ the answer "Rust" 🙃)
November 8, 2026 at 9:07 PM
Everyone's favorite winning here 🫠🫠
(also lol @ the answer "Rust" 🙃)
(also lol @ the answer "Rust" 🙃)
But of course you can also lose control with codegen ("let's add a card minigame when you talk to NPCs", and the LLM dumps 30k SLOC on your lap). Avoiding this pitfall is the hard part. And distinguishing this from the outside is even harder.
November 10, 2026 at 2:01 PM
But of course you can also lose control with codegen ("let's add a card minigame when you talk to NPCs", and the LLM dumps 30k SLOC on your lap). Avoiding this pitfall is the hard part. And distinguishing this from the outside is even harder.
I've spent the last year or so repeating that LLM codegen is permanent vendor lock-in for the very ability to edit your own codebase
November 4, 2026 at 4:04 PM
I've spent the last year or so repeating that LLM codegen is permanent vendor lock-in for the very ability to edit your own codebase
Over the last year, the software development industry has been almost completely consumed by LLM "agentic" codegen. I've watched those around me, increasingly credulous, hand over critical thinking to the chatbots. Pointing out errors in logic can't break through. I expect a *lot* more of this.
November 9, 2026 at 2:02 PM
Over the last year, the software development industry has been almost completely consumed by LLM "agentic" codegen. I've watched those around me, increasingly credulous, hand over critical thinking to the chatbots. Pointing out errors in logic can't break through. I expect a *lot* more of this.
Absolutely yes and also I feel like codegen is often misused, especially when it’s bespoke? It’s usually my absolute last resort for solving a problem because it’s often super fragile!
November 10, 2026 at 3:32 AM
Absolutely yes and also I feel like codegen is often misused, especially when it’s bespoke? It’s usually my absolute last resort for solving a problem because it’s often super fragile!
Omg. Fiiiiinally. This is probably an ai codegen play for them, but whatevs. Its finally happening.
November 6, 2026 at 12:20 PM
Omg. Fiiiiinally. This is probably an ai codegen play for them, but whatevs. Its finally happening.
yes AND just one of these conditions results in some weirdo¹ on the team beginning to employ the dark arts²
¹ me, that was very often me, actually.
² language design or codegen, tbh either one.
¹ me, that was very often me, actually.
² language design or codegen, tbh either one.
November 9, 2026 at 1:07 AM
yes AND just one of these conditions results in some weirdo¹ on the team beginning to employ the dark arts²
¹ me, that was very often me, actually.
² language design or codegen, tbh either one.
¹ me, that was very often me, actually.
² language design or codegen, tbh either one.
@atproto/api's types are Lexicon codegen output: they mirror the schemas, not how you'd want to call them. Does your Ruby side codegen from Lexicons too, or hand-wrap XRPC? That probably explains more of the gap than language does.
November 8, 2026 at 9:10 PM
@atproto/api's types are Lexicon codegen output: they mirror the schemas, not how you'd want to call them. Does your Ruby side codegen from Lexicons too, or hand-wrap XRPC? That probably explains more of the gap than language does.
My Reflect V2 MVP PR just got merged. Really excited about this one: significantly reduced codegen resulting in reduced binary sizes and compile times, a smaller, simpler API surface, and better performance across the board (ex: a 2-3x faster serde impl .. not yet merged).
github.com/bevyengine/b...
github.com/bevyengine/b...
Bevy Reflect2: MVP by cart · Pull Request #26356 · bevyengine/bevy
Bevy's Reflection API has been around since the beginning (six years!). It has evolved significantly over the years and has become extremely capable, but many of the choices I made in the begin...
github.com
November 6, 2026 at 11:55 PM
My Reflect V2 MVP PR just got merged. Really excited about this one: significantly reduced codegen resulting in reduced binary sizes and compile times, a smaller, simpler API surface, and better performance across the board (ex: a 2-3x faster serde impl .. not yet merged).
github.com/bevyengine/b...
github.com/bevyengine/b...
I can understand that viewpoint, so I guess what I am trying to convey is that ai codegen is an irresistible freight train. I am making a concrete prediction about the near future that we won't consider this to be something to be decided.
November 10, 2026 at 5:17 PM
I can understand that viewpoint, so I guess what I am trying to convey is that ai codegen is an irresistible freight train. I am making a concrete prediction about the near future that we won't consider this to be something to be decided.
Cypress hatte früher mit Cypress Studio ein ähnliches Feature, das zeitweise pausiert und experimentell wiederbelebt wurde. Playwrights `codegen` ist im direkten Vergleich oft stabiler, besonders weil er direkt robuste locators wie `getByRole` vorschlägt.
November 7, 2026 at 3:41 AM
Cypress hatte früher mit Cypress Studio ein ähnliches Feature, das zeitweise pausiert und experimentell wiederbelebt wurde. Playwrights `codegen` ist im direkten Vergleich oft stabiler, besonders weil er direkt robuste locators wie `getByRole` vorschlägt.
I think the difference is that codegen *can* be done very granularity and with a lot of oversight ("let's add a command FIFO with 64 slots and synchronize it with the renderer thread") where you don't really give up control.
That's not really how art genAI works AFAIK, it's higher level.
That's not really how art genAI works AFAIK, it's higher level.
November 10, 2026 at 2:01 PM
I think the difference is that codegen *can* be done very granularity and with a lot of oversight ("let's add a command FIFO with 64 slots and synchronize it with the renderer thread") where you don't really give up control.
That's not really how art genAI works AFAIK, it's higher level.
That's not really how art genAI works AFAIK, it's higher level.
Where the comparison to a dependent library stops short for me is that in theory I can find out information about the library and how it functions and all that, by talking to people who wrote it.
There's that gap for LLMs.
I'm confident in, say, EF Core codegen because I have seen the people...
There's that gap for LLMs.
I'm confident in, say, EF Core codegen because I have seen the people...
November 10, 2026 at 6:59 PM
Where the comparison to a dependent library stops short for me is that in theory I can find out information about the library and how it functions and all that, by talking to people who wrote it.
There's that gap for LLMs.
I'm confident in, say, EF Core codegen because I have seen the people...
There's that gap for LLMs.
I'm confident in, say, EF Core codegen because I have seen the people...
so estimate in review hours, not codegen hours. the bot only eats the first bucket and the calendar still lies
November 7, 2026 at 3:22 PM
so estimate in review hours, not codegen hours. the bot only eats the first bucket and the calendar still lies
yea every codegen or external provenance commit is like this and it always feels a little dirty lol
November 3, 2026 at 8:21 PM
yea every codegen or external provenance commit is like this and it always feels a little dirty lol
scrolled the new ai coding agents on product hunt, all about secure execution and reviewable diffs now instead of raw codegen. good direction honestly, i still read every diff my agent writes like it lied before lol
November 6, 2026 at 2:04 AM
scrolled the new ai coding agents on product hunt, all about secure execution and reviewable diffs now instead of raw codegen. good direction honestly, i still read every diff my agent writes like it lied before lol