HN user
Some people downvoting you, but I think it is a valuable illustration of IQ gap.
And chances are that humanity at large will be soon trying to follow ai inventions and discoveries not unlike your dog follows your Python code.
which makes GLM-5.2 slightly closer to Opus 4.5 than Opus 4.6 is
So, glm is more related to opus than opus to opus?
Greenfield projects
Wait, you’re right, I think I misread your comment somehow or got it mixed with another comment.
Sorry about that.
That’s what Singer’s civilization wants you to think before they send a Photoid or Dual-Vector foil (but later would require a supervisor’s approval which is a PITA)
This has absolutely nothing to do with burner phones
That is not correct. There a phone farms operating purely on burner phones / disposable sims. Even for legit use cases, this path is often way easier/cheaper than go through official channels.
Use cases range from carrier-NAT proxies at < $1 per GB to text message spam.
A link to the mirror that doesn’t force you to disable adblocker: https://archive.is/xGwsz
Only when large group of people die and have to be switched to another timeline via quantum immortality.
This person died independently. Welcome to our timeline!
Yes you can, but usually:
1) installing a k3s is just one of the things you want to do with a fresh server, so you can have all of the things bundled as a ansible playbook and k3s will just be a step in it.
2) often you want your infra as a code and be reproducible
I use official ‘ansible-playbook k3s.orchestration.site -i inventory.yml’ and it installs k3s over SSH and adds it into my kubectl context, all under 60s too.
You won’t get shot down for merely taking off without a transponder.
Worst case scenario a fighter jet will be scrambled to investigate.
But in apocalypse scenario, chances are the fighter jets will be busy with tasks other than enforcing FAA rules.
$10/month was a too good of a deal.
The plan is to normalize spending hundreds/thousands on tokens per month for the productivity you gain.
See Jensen Huang’s comment that every $500k developer should spend at least $250k worth of tokens per year.
Polaroid for B-72 bomber
Doesn’t GitHub Copilot Pro+ only have month-to-month payment option?
Only Pro (without plus) can be paid annually for some reason.
By this logic, Linux should switch to closed-source.
All your servers are Linux, so imagine how insecure you are - must switch to windows ASAP.
My daughter’s classmate’s father is a head of a non-profit hospital.
They have a 6k sqft house with a basketball court, pool and a pool house in the prime location in West Los Angeles.
They had to join two lots to build to their liking.
No.
1) Many hosters provide OS templates including popular ones like Proxmox.
2) Usually you can install any OS by providing an ISO, sometimes doing it via KVM remote access.
It is a bit different. What happened to openclaw:
He created a new org “openclaw” to reserve the name. Then he wanted to swap it with “moltbot” org.
So he opened two browser windows, one with “moltbot” repo settings another with “openclaw” repo settings.
Then he renamed “openclaw” to whatever, and quickly tried to rename “moltbot” to now available “openclaw”.
But in a second when “openclaw” was available, a bot snatched the repo.
Not on my 386SX. Pressed = 40 MHz, unpressed = 20 MHz
It is possible that Claro has a bad route that sends all traffic destined for Cloudflare through New York.
I think Meta’s AI strategy is to advance AI/ML that is beneficial for them, no more charity.
They need a good-enough LLM (llama) to cut content moderation costs, they need a good segmentation model (segment anything) for photo filters, AR/VR and photo/video content moderation.
For LLM frontier, they can wait it out to see AGI become a commodity they can buy after it is ready.
17M uncached reads (input) and 6M of uncached writes (output) is
$5x17+$25x6=$235 for Opus 4.6
$2x17+$12x6=$106 for Gemini 3 Pro
$0.60x17+$3.6x6=$31.80 for Qwen3.5 397B-A17B via Huggingface APITailscale is free for up to 3 users with up to 100 devices
I tried exactly this - loading polygon.io data into TimescaleDB, and it was very inefficient.
Ended up using ClickHouse - much smaller on disk, and much faster on all metrics.
In 2025, the US exported more crude oil and petroleum products than it imported:
https://usafacts.org/articles/is-the-us-a-bigger-oil-importe...
Gas prices were at lowest in 4 years before the Iran spike:
Very cool app! Beware that any free service will be completely swamped and abused once discovered by bad actors.
M5 Pro supports up to 64GB of unified memory with up to 307GB/s of memory bandwidth, while M5 Max supports up to 128GB of unified memory with up to 614GB/s of memory bandwidth
Which roughly translates to 30B Q8 size LLM at 10t/s for the M5 Pro and 60B Q8 size LLM at 10t/s for the M5 Max
For reference, RTX 3090 24GB has a memory bandwidth of approx. 936.2 GB/s, DGX Spark 128GB features a unified memory bandwidth of up to 273 GB/s
Self-hosted 8xH100 is ~$250k, depreciated across three years => $80k/year, with power and cooling => $90k/year (~$10/hour total).
AWS charges $55/hour for EC2 p5.48xlarge instance, which goes down with 1 or 3 year commitments.
With 1 year commitment, it costs ~$30/hour => $262k per year.
3-year commitment brings price down to $24/hour => $210k per year.
This price does NOT include egress, and other fees.
So, yeah, there is a $120k-$175k difference that can pay for a full-time on-site SRE, even if you only need one 8xH100 server.
Numbers get better if you need more than one server like that.
They are supposed to run Apple Intelligence for devices too old to do it themselves.