veunes

7 ms·

121.
by veunes 8mo ago
He had the full source code of a working Linux driver that does exactly the same thing, just in a neighboring kernel dialect. The task was to translate, not invent. Sure, it's still impressive (given the difference in kernel APIs), but
122.
by veunes 8mo ago
To be fair if you open up driver source code from the vendors themselves, it's often the same hell with magic numbers and lack of checks because "we know what the hardware will return". But you're right on the main point
123.
by veunes 8mo ago
I wouldn't call this "clean-room". The models were trained on all available open source, including that exact original Linux driver. Splitting sessions saves you from direct copy-paste in the current context window, but the w
124.
by veunes 8mo ago
Spot on about keeping that AGENTS.md and logging all decisions. Letting an agent code for a long stretch without pinning down the state is a surefire way to end up with a Frankenstein codebase. Forcing it to document why you ditched LinuxKP
125.
by veunes 8mo ago
Active account with real interactions = more normal. Which is a pretty telling product story in itself
126.
by veunes 8mo ago
The part I really agree with is the social impact
127.
by veunes 8mo ago
Once a group gets big enough...
128.
by veunes 8mo ago
Your mum's experience is probably what FB is best at: high-trust network, lots of original photos, lots of comments from real friends
129.
by veunes 8mo ago
The depressing part is that generative slop is a perfect match for the incentive structure: infinite supply, tuned to trigger comments, and you don't need real creators. If your product metric is time-on-site, this is what you get
130.
by veunes 8mo ago
My guess is we'll see a gradual slope rather than a cliff
131.
by veunes 8mo ago
Ironically, the fact that fully autonomous systems are more efficient when feasible is exactly why the exoskeleton analogy makes sense
132.
by veunes 8mo ago
Communication overhead between humans is real, but it's not just inefficiency, it's also where a lot of the problem-finding happens. Many of the biggest failures I've seen weren't because nobody could type the code fast
133.
by veunes 8mo ago
In the self-driving case, the safety driver often isn't contributing much to the system's performance
134.
by veunes 8mo ago
What's interesting to me is that most real productivity gains I've seen with AI come from this middle ground: not autonomy, not just tooling, but something closer to "interactive delegation"
135.
by veunes 8mo ago
Migrating off Actions is a pain, but the real lock-in is the network effect. Moving to something like Codeberg means losing all those "drive-by" contributors who won't bother signing up for a new instance just for one small f
136.
by veunes 8mo ago
The asymmetry is brutal: generating slop costs pennies and five seconds, while reviewing it costs senior engineers hours. The natural filter of complexity is broken, so the only way out is to artificially reintroduce friction Projects like
137.
by veunes 8mo ago
You don't need frameworks for one-off scripts, but in prod, you're going to need RAG, proper memory, tools, and orchestration anyway. Without standards you'll just end up writing your own janky framework on top of requests. L
138.
by veunes 8mo ago
Spot on. BEAM is great at surviving process crashes, but if the whole cluster goes down or you redeploy, that in-memory state evaporates. It's not magic. For agents that might hang around for days, pure Elixir isn't enough, you st
139.
by veunes 8mo ago
The Let it crash concept is perfect for deterministic bugs, but does it work for probabilistic errors? If an LLM returns garbage, restarting the process (agent) with the same prompt and temperature 0 yields the same garbage. An Erlang Super
140.
by veunes 8mo ago
The problem isn't the abstraction level, it's the loss of the incubation period. Moving from Assembly to C didn't remove the need to think through data structures. Prompt engineering, however, triggers skipping the reflection
141.
by veunes 8mo ago
That's the hidden price of fast development
142.
by veunes 8mo ago
Validation is always harder than generation. Writing yourself means building context brick by brick. Reviewing AI means reconstructing someone else's - often broken - logic from zero. Juggling three projects in parallel fragmentizes me
143.
by veunes 8mo ago
My point wasn't that it's impossible, but that it creates a massive layer of overhead. Previously the project "state" lived in the developer's head; now for the agent to be effective, you have to constantly offload
144.
by veunes 8mo ago
Character matters but so does having people around you who are willing to call it early, before you've rationalized yourself into ignoring it
145.
by veunes 8mo ago
But less about personal brilliance and more about how social power actually works when money, status, and weak accountability intersect
146.
by veunes 8mo ago
But it might've changed one decision, one meeting, one normalization step
147.
by veunes 8mo ago
I'd rephrase it as: nobody should be trusted with unchecked power, especially when it's exercised quietly and indirectly
148.
by veunes 8mo ago
Sometimes (sometimes) it just implies that someone sent an email, got ignored, and left a paper trail behind
149.
by veunes 8mo ago
All of this speedrun hits a wall at the context window. As long as the project fits into 200k tokens, you’re flying. The moment it outgrows that, productivity doesn’t drop by 20% - it drops to zero. You start spending hours explaining to th
150.
by veunes 8mo ago
This is the birth of Shadow AI, and it’s going to be bigger than Shadow IT ever was in the 2000s Back then, employees were secretly installing Excel macros and Dropbox just to get work done faster. Now they’re quietly running Claude Code in

More ›