HNHacker News
TopNewBestAskShowJobs

ElenaDaibunny

32 karma · joined April 14, 2026

ML engineer, mostly doing on-device inference and agent stuff on Apple Silicon. Python / Swift / MLX.
submissionscomments
ElenaDaibunny··on Cyberpunk Comics, Manga and Graphic Novels
a24 style really hits
ElenaDaibunny··on GitLost: We Tricked GitHub's AI Agent into Leaking Private Repos
Additionally did all that? man
ElenaDaibunny··on Canada's only watchmaking school still ticking after 80 years
yep,sci-fi really got this one wrong lol.
ElenaDaibunny··on Canada's only watchmaking school still ticking after 80 years
kid me thought AI would replace people doing insanely precise hand work. turns out it replaced me writing emails, and this school's still going strong after 80 years. lmao we got played
ElenaDaibunny··on How to sequence your own DNA at home
just a hobby project for now,pretty wild that this can be done at home.
ElenaDaibunny··on Open Source Low Tech
this is the kind of open source project that makes me question my life choices every time i install a node module
ElenaDaibunny··on Tokenmaxxing is dead, long live tokenmaxxing
i think tokenmaxxing mostly comes from cloud pricing. once you're paying by the token, you naturally start caring about token counts. with local inference i barely think about it anymore.
ElenaDaibunny··on Show HN: Graphical SQL Builder and Debugger
visual SQL builders have been around forever, most feel like they were designed by someone who never writes joins, this one has reverse-engineering from existing SQL which is enough to try it
ElenaDaibunny··on Qwen-AgentWorld: Language World Models for General Agents
10M trajectories, probably more of a data scale win than a world model breakthrough tbh
ElenaDaibunny··on How I play video games with spinal muscular atrophy
curious how playability handles the false positive problem with expressions though, involuntary facial movements must trigger inputs constantly
ElenaDaibunny··on Lisp in the Rust Type System
call/ec at the type level is the surprising part here, most compile-time Lisp implementations stop at basic eval but escape continuations need real control flow reasoning through trait resolution
ElenaDaibunny··on Midjourney Medical
spa as a regulatory bypass is clever, body comp data first and diagnostics later. 500k transducers doing full body ultrasound in 60s is a massive hardware bet for an image gen company tho
ElenaDaibunny··on DOJ claims xAI's gas turbines are a matter of 'national and energy security'
the most creative regulatory arbitrage I've seen this year,haha
ElenaDaibunny··on Ask HN: Has anyone replaced Claude/GPT with a local model for daily coding?
we've been building local agents with vision models, works great for gui automation but coding tasks still need cloud models for reliability
ElenaDaibunny··on Understanding the rationale behind a rule when trying to circumvent it
of course they had to add dont wait for your worker thread in 2020
ElenaDaibunny··on DeepSeek V4 Pro beats GPT-5.5 Pro on precision
Yep, matches my experience. gpt keeps adding fields and changing types on structured output when you need it to just follow the spec~
ElenaDaibunny··on C++: The Documentary
wild that c++ is apparently the fastest growing top 4 language right now
ElenaDaibunny··on The ways we contain Claude across products
Egress controls are the only real defense here and most people running Claude Code locally dont have any.
ElenaDaibunny··on AI Engineers aren't safe from being replaced by AI
The people who stick around will be the ones who deeply understand the problem, not the ones who are good at wrangling the tools.
ElenaDaibunny··on Florida sues OpenAI and Sam Altman over AI risks
if state level ai lawsuits become a trend the compliance costs alone will push smaller players out of the market. openai can afford lawyers, most startups cant
ElenaDaibunny··on OpenAI frontier models and Codex are now available on AWS
was only a matter of time. enterprise teams on aws werent going to rearchitect their stack just for model access, easier to bring the models to where the workloads already are
ElenaDaibunny··on Microsoft data suggests using AI is more expensive than hiring people
token prices are dropping fast though so this comparison has a short shelf life
ElenaDaibunny··on Rust (and Slint) on a Jailbroken Kindle
zig as a cross-compile backend for rust is so good.
ElenaDaibunny··on Google employee charged with $1M Polymarket insider trading bet on search term
Forbes flagged this account back in December and it still took prosecutors months to charge him.
ElenaDaibunny··on Flatpak Will Depend on Systemd
At this point the Linux desktop stack has a harder systemd dependency than most people realize, Flatpak was one of the last holdouts.
ElenaDaibunny··on Browser-based file encryption tool using WebCrypto
Having both PBKDF2 and Argon2id as options is nice, most similar tools only bother with PBKDF2.
ElenaDaibunny··on Microsoft Copilot Cowork Exfiltrates Files
Every new integration is another exfil surface, this was bound to happen.
ElenaDaibunny··on Using AI to write better code more slowly
The bug-finding use case alone makes this worth it.
ElenaDaibunny··on DeepSeek reasonix, DeepSeek native coding agent with high caching and low cost
The caching strategy is doing most of the heavy lifting here cost-wise.
ElenaDaibunny··on Memory has grown to nearly two-thirds of AI chip component costs
unified memory architectures are getting more interesting for inference workloads.
Page 1 of 2Next →