lima

It's a competitive market. Kimi K3 has almost reached parity with Opus.

It always worked like that

Even if you can't authenticate at the application level, it's still much better to encrypt/authenticate traffic on the wire (using a VPN or something like Tailscale) instead of 802.1x auth.

Non commercial licenses are not generally considered open source

Google solved credential phishing a long time ago using hardware tokens (gnubby - the predecessor to FIDO/U2F/Passkeys...).

There's other types of social engineering, but phishing is mostly an engineering issue.

Probably 802.1x, but it's easy to bypass if you have access to an authorized device. This kind of authentication has to be done at the application level, treating the network as a perimeter doesn't work.

The company also should have restricted network access to the port in the conference room so that an unknown device like a Raspberry Pi could not make an Ethernet connection from that spot

Bad take - the actual problem is that there was a trusted network in the first place. This kind of network access control is trivial to bypass, and trusted devices can get compromised.

This. It's easy to forget that Postgres is fundamentally a single-node database without distributed transactions. It won't pass the Jepsen test suite with multiple nodes. DBOS, Temporal and friends inherit this limitation.

Something like Restate actually implements distributed transactions.

Gerrit's new UI comes pretty close, and it's also what Google uses for projects living outside of google3.

GitHub's PR review workflow is archaic in comparison, especially now that every page takes 2-3 seconds to load.

Don't worry, they accept PRs on that repo. They just merge them internally and then re-export them.

There are some variations, but this is generally the same with all open source projects which live in their internal monorepo, such as gVisor or Bazel.

Yes it will, there's a clear purpose and the customer explicitly agrees.

Fable on GCP requires accepting a 60-day retention policy: https://cloud.google.com/terms/advanced-ai-safety-addendum

I don't think it mentions sharing the data with third parties such as Anthropic?

Higher IO latencies in HDs might actually make this attack easier - more contention means more bits of data.

Shor published multiple quantum algorithms, including one for discrete logarithms. The term is sometimes used interchangeably.

They're closely related, ECC and RSA are both instances of the hidden subgroup problem.

How do you know the clanker respects the instruction not to search the internet?

You can't, but given that it's a previously unsolved problem, it doesn't seem relevant? (nor are the author's potential biases - the claims are easily verified independently)

We tried this, but the quota for Opus models defaults to 0 on VertexAI and quota increase requests are auto-rejected.

Any tips?

No, unless you count tricks which are explicitly against ToS

They use a buffer battery, it's quite feasible with that.

You can still use OpenCode with the Anthropic API.

Fun fact: In Germany, the civil courts will usually take the case anyways if it has merit, but the winner ends up paying for the whole lawsuit if they failed to make an effort to resolve the case before suing.

And Hungary is pretty much a rogue EU state - their government did go full authoritarian and is aligned with Russia.

It's true, neither AWS nor GCP support spending limits. Only alerting.

Everything gets more expensive?

How does this approach compare to the various Ghidra MCP servers?

But the implementation of Gerrit seems rather unloved

There are lots of people who are very fond of Gerrit, and if anything, upstream development has picked up recently. Google has an increasing amount of production code that lives outside of their monorepo, and all of those teams use Gerrit. They have a few internal plugins, but most improvements are released as part of the upstream project.

My company has been using it for years, it's a big and sustained productivity win once everyone is past the learning curve.

Gerritforge[0] offers commercial support and runs a very stable public instance, GerritHub. I'm not affiliated with them and not a customer, but I talk to them a lot on the Gerrit Discord server and they're awesome.

[0] https://www.gerritforge.com/

Clearly, convincing it otherwise is part of the challenge.

This is changing, Ghidra is increasingly replacing IDA for commercial work.

LLMs don't really know why they got something wrong, so unless it had access to the original chain of thought, it's just guessing.

It's really the same thing. Network partitions can (and do) occur within a single DC.

If your network is down, your clients can't connect DB too, and nothing works

This sounds simple, but can only be achieved with high certainty using distributed consensus.

It's complex, but Ceph's storage and consensus layer is battle-tested and a much more solid foundation for serious use. Just make sure that your nodes don't run full!