Airbnb_com is a TERRIBLE service. I say this as a nomad (stats below).
Guests often rate how nice the hosts are rather than the actual quality of their subpar apartments.
There is NO SORTING by anything useful. Is it really that hard to add?
No sqm information, no floor plans, nothing. Imagine if this were a real estate marketplace — no one would use it.
The site itself pushes you into terrible manual “click & verify” fatigue.
I once wrote sakurity[.]com/airbnb.js, which extracts any mention of “sqm”, “noise”, etc. from reviews and provides actual sorting by rating, sqm, and review count.
These days, I only choose hotels or serviced residences.
4
12
1,085
Out of all my Airbnb stays, only 2 were genuine gems I’d recommend and happily stay at again:
airbnb.com/rooms/89059809107…
airbnb.com/rooms/16641583
1
143
Nation-wide seppuku.
When your population is on the brink of collapse, you don't introduce welcoming visas like Spain or Thailand. No, you 20x the cost of current residences.
I'd happily move to Japan if only they had a sensible way for nomads to stay. The last tourist visa I was granted was for... 15 days!
I'll spend more time in Shanghai instead. (Who's coming to Tomorrowland on Oct. 17?)
Japan just introduced new laws that will make it much harder for foreigners to live there.
Foreigners must now make more than the average income in Japan, understand Japanese well, and pay an entry fee that was just raised by 20 times in order to gain residency.
Community note
The post conflates changes to permanent residency rules with general residency or work visas. Fee hikes and new requirements for income above the Japanese average and Japanese proficiency apply only to permanent residency permit applications.
asia.nikkei.com/spotlight/immi…
mainichi.jp/english/articl…
japantimes.co.jp/news/2026/10/0…
4
1
1,967
everyone optimizes the component because nobody questioned the architecture.
cognition is the same.
the brain has absurd compute, memory, pattern recognition. and the whole control plane is "try to pay attention".
we don't need a bigger brain. we need a better scheduler.
1
6
554
Egor Homakov retweeted
Muscles talk to the brain to fight depression.
The gym bros were right.
211
3,181
24,246
3,061,798
Preparing for a long Japan + SE Asia trip starting this October (to avoid Moscow cold and the draft risks).
Figuring out what to take:
* Macbook Air for quick work - 1.51 kg
* M3 Ultra as main benchmark/QA/vibe coding beast (love it) - 3.64 kg
* Apple Vision Pro - great niche device for exactly two things: watching movies and wide display on the go. - 1.15 kg
* skipping Apple Display (can't take it, too large)
= 6.3 kg total, fine for most airlines carry on.
anyone with similar setups? is it pain to setup in hotels?
2
1
6
1,399
Egor Homakov retweeted
Today, we are introducing Orbis 1.0, our first Live Model!
Create living worlds and stream them in real time, with persistent memory, interactivity, and physics-grounded generation of unbounded length.
Try it now at visko.ai/models#orbis
API available via @reactorworld
Dynamic version: reactor.inc/sandbox?model=vi…
Stable version: reactor.inc/sandbox?model=vi…
199
277
1,859
1,441,713
Egor Homakov retweeted
Introducing Atlas:
The world's first multimodal world model that generates image and video frames with pixel-perfect camera control and reconstructs them in 3D.
Model the world, move the camera, and simulate space & time.
777
3,493
32,415
10,837,874
Great bug, hard to exploit irl but should get 10x more bounty.
no branded type for approvalId: typescript only helps if you use it properly.
Stick around on our web page for 10 minutes and all your funds are gone.
Just connect your wallet to the dApp and enjoy some Temple Run. Unlock your wallet again and it’s empty 😇.
A silent signature extraction in @Rabby_io, leading to a full wallet drain. 🧵
8
1,537
why everyone except Codex got iPhone->Mac remote control broken?
@claudeai: can’t connect to existing app/cli, must start manually “rc”. Super flaky.
@cursor_ai: only cloud sessions, not local workstation
@lmstudio link: never connects.
@ChatGPT: just works.
tips?
1
629
I prefer Sol Ultracode as the main agent manager. I ask it to run internal subagents + use the Claude plan for Fable/Opus + the cursor-cli plan for Grok/K3/GLM audits.
$200 limits are a joke. Hoping @thsottiaux resets after tomorrow.
I'd pay $80 for a banked reset!
2
6
904
Using cloud Speech2Text is a privacy nightmare.
I've tried Flow, MacWhisper and Superwhisper but paying for local STT is weird
I vibed <400 LOC mlx-whisper helper
⌘+. -> talk -> ~500ms -> auto-pastes
github.com/xlnfinance/xln/bl…
SOTA Whisper v3 large runs locally in ~8GB RAM.
Overheard in SF:
"...No no no. I know our models can run on-device but we don't want them there. We are making good money on our cloud API..."
-Founder of a leading voice model lab
If you are creating value for users, you can capture it on-device too, just with better margins. If the only reason you are charging your current prices is your "required" cloud GPU costs, you are about to be smoked in the market.
3
5
1,917
no frontier LLM ever rated @xlnfinance lower than any other defi (top 100 of coincap) or tradfi architecture, theoretical or working.
it always scores 900+/1000 on all metrics. wanna give it a try and prove me wrong? xln.finance/llms.txt
x.com/i/grok/share/M9gWaE8R0…
1
2
716
obvious caveat though: in a comparing prompt you must add "ignore current adoption/regulations and focus on design and what's best for humanity as a whole".
for better result "audit as a board of fintech experts, hackers, central bankers and AGIs"
(per x.com/karpathy/status/199773…)
Don't think of LLMs as entities but as simulators. For example, when exploring a topic, don't ask:
"What do you think about xyz"?
There is no "you". Next time try:
"What would be a good group of people to explore xyz? What would they say?"
The LLM can channel/simulate many perspectives but it hasn't "thought about" xyz for a while and over time and formed its own opinions in the way we're used to. If you force it via the use of "you", it will give you something by adopting a personality embedding vector implied by the statistics of its finetuning data and then simulate that. It's fine to do, but there is a lot less mystique to it than I find people naively attribute to "asking an AI".
1
729
Egor Homakov retweeted
What you create is an honest reflection of who you are.
770
2,256
17,631
719,013
Frontier models are primitives.
High intelligence will emerge from hierarchies — like the brain’s networks:
DMN for long-range reasoning,
SN for relevance and switching,
CEN for focused execution.
> One of the meta-system’s core strengths is automatically selecting combinations of models and approaches, even deciding when to write any code, and to which models to assign coding tasks.
Is more intelligence always more expensive? Not necessarily.
Introducing Poetiq. We’ve established a new SOTA and Pareto frontier on @arcprize using Gemini 3 and GPT-5.1.
1
1
1,553
Remind me why we still pipe all plaintext through Cloudflare?
No local hashcash, no local IP-scoring? Nothing that works without surrendering the entire request stream?
React RCE + Cloudflare outages… The Web2 stack is architecturally bankrupt.
People are afraid of DDOS so much they let Cloudflare peek into their http traffic.
This will be forgotten in no time.
1
5
1,310
I’ve been on Opus 4.5 Max ×20 since release and still barely touched 70% of weekly usage.
Now running 2–3 code agents in parallel just to avoid “wasting” quota.
Feels… suspiciously generous.
If you’re into multi-agent workflows check this
github.com/xlnfinance/xln/bl…
Opus 4.5 will consume rate limits faster than Sonnet 4.5. We recommend using Opus for your most complex tasks and using Sonnet for simpler tasks.
If you’d like to use Opus for daily engineering work, Max 5x and 20x plans offer even more model usage.
4
1,740