This week Max taught the whole company how 100+ servers talk to each other without the public internet ever finding out.
Wisdom Wednesday is where one person at @byteful teaches everyone else something they know well. Max's was @Tailscale. One private network across datacentres, ISPs and cloud providers, nothing left open to the public web, and when a traffic rule changes you change it once instead of touching every machine.
Open source and written in Go, which visibly pleased Max more than the security did.
(We are still hiring a full-time Support Engineer to join us in person in Manchester. There will be more useful nerdiness.)
Consistently best😤 @Byteful is coming out fastest on average across @AIMultiple's residential benchmark
- fastest on 9 of the 18 days
- never dropped out of the top 3
They rerun it every 5 minutes against amazon, bing, ebay and youtube, so there’s no benchmark day to prepare for.
Three things to check when a proxy request fails 🧰
1. The response headers: status code and request ID
2. The Proxy User: credentials, access and limits
3. Live Activity: the request, its error and target
Check these before changing the setup.
A monthly IP count and a live pool count measure different things 📊
- Monthly active IPs: every unique address seen across a month.
- Live pool: nodes online right now that match your request.
The first shows how broad a network is, the second what your job can draw on today.
Google's had two goes at this now!
The same issue both times: what copyrighted content is SearchGuard actually protecting, and who authorised Google to protect it?
A good win for the open web
A big win for SerpApi and everyone who believes in an open internet.
For the second time, a federal court has dismissed Google’s claims against SerpApi, granting our motion to dismiss its amended complaint.
This is an important development for our company, our customers, and our continued commitment to making publicly available search data accessible.
Our General Counsel, Chad Anson, breaks down the court’s latest decision and what comes next.
Read the full update: serpapi.com/blog/court-dismi…
HMRC thinks only ~600 people a year will qualify
my read is the high bar is kind of the point
10% CGT on up to £10m, but only if you've actually built and scaled the company
if it gives founders another reason to keep hiring + growing instead of selling early, that's a good incentive
We want the best entrepreneurs to start, scale and stay in Britain.
That’s why we’re backing businesses at every stage of growth, including with a new Founders’ Relief to encourage founders to scale here and reinvest in the next British success story.
A script that generates proxies doesn't need to be able to buy anything…
every API action can be scoped separately now, with an expiry if you want one
matters a lot more once it's an agent holding the key
Give an automation only the API access it needs 🔑
Byteful API actions can be scoped individually. A script that generates proxies gets a key that only generates proxies, with an expiry date if you want one.
If a key leaks, the damage stops where its permissions do.
🌐 Can a Rust-powered TCP/IP stack look like Windows, Linux, or Apple?
Joe Dye is about to take the stage at #RustLab2026!
Discover how user-space networking enables fingerprint spoofing at scale, handling thousands of requests per second! 🦀
Request count × page size is a rough guess at best…
retries, rendering and payload all move the number
measure bytes per usable result on a real sample and size from that
Measure the workload before you buy data 📊
Counting requests gives a rough guess. Measuring data per successful result is what to size a tier on, since payload, rendering and retries are already in it.
Residential and mobile data does not expire on Byteful.
One-region proxy tests miss a lot…
same proxy
same target
different region
different result
We built the tester to run US + Europe + Asia at the same time. Works with all providers, not just us.
Test a proxy from where you will actually use it 🌍
Our proxy tester runs the same proxies from servers across the US, Europe and Asia at once, against a target you choose. Results per region, with status codes.
Works on any provider's proxies. Free, no account needed.
mobile targeting only helps if the supply is there
pick:
> country
> region
> city
> carrier
we show how many mobile nodes match right now
6M+ is the total network
live count is what you can actually route through
same data is in the API too
Byteful shows how many mobile nodes match before you generate anything 📍
Pick a country, region, city or carrier and the dashboard updates a live count of nodes currently online and matching.
Same data in the public API, so an automated job can check before it runs.
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.
Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.
Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.
But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?
We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.
Read the full report: cohere.com/blog/concerning-b…
Bot detection keeps moving down the stack…
user agents
JS fingerprints
TLS
now TCP
Before HTTP, the handshake can already leak the OS behind it so we rebuilt TCP in user space for ISP & Datacenter proxies.
Try the Beta here:
byteful.com/blog/introducing…
Formatting a proxy list by hand is the sort of small job that eats twenty minutes and teaches you nothing... one missing colon and the import fails. Free formatter, whole list at once. It won't tell you which proxies are live, just how the list is written.
One missing colon can stop a proxy list importing 🔧
Our proxy formatter converts a whole list at once. Paste or upload, it works out the source format on its own, you pick the output, then copy or export.
It rewrites the list. It will not tell you which proxies are live.