Tool calls (particularly fetching for context) eats the context window heavily. I explicitly send MCP calls to sub agents because they are so “wordy”.

Everyone who has not hit this bug thinks it’s user error… It’s not. It happened to me a few days ago, and the speed at which I tore through my 5 hour usage cap was easily 10x faster than normal.

Also: sub agents do not get you free usage. They just protect your main context window.

I'm on Max. This morning, just to test, before doing anything else whatsoever, I was at 0%, and I typed 'test one two three' into CC.

That put me at 12%.

I have no MCPs except the built in claude-in-chrome.

This is clearly a bug.

Don't they consume less of the token quota in case the subagents are running cheaper models like Sonnet and Haiku compared to Opus?

Correct—I just wouldn't want folks to mistakenly think that the context fill % corresponds 1:1 with session token use.

Yes, sorry. I meant it more as a descriptor of how many tokens it consumes. You are still stuck burning money.

Readimg through this thread, it seems likely is a KV cache "bug". Theyre likely doing too many evictions of the LLM cache so the context is being reloaded to often.

Its a "bug" because its probably an intended effect of capturing the costs of compute but surfacing a fact that they oversold compute to a situations where they cant keep the KV cache hot and now its thrashing.

Caching helps them too, so I hope they fix it

In the past it had less to do with seizing the vessels and more to do with keeping financial flows between organizations offering shipping services and oil hidden from the banking system. America could have easily seized any ship they wanted to during the sanctions over the past decade. They didnt because the sanctions are American constructs: they dont apply on the open seas where UNCLOS matters. America can still seize them, but the legality is murky and comes with a reputational cost.

Now with Hormuz closed, America needs every last oil barrel moving so the economy doesn’t grind to a halt. Remember, it’s a war of choice for the US. We don’t need Iran gone as much as we want low oil prices.

> the sanctions are American constructs: they dont apply on the open seas where UNCLOS matters

Technically correct. But the way these countries evade U.S. sanctions is by flying false or no flag. That, in turn, makes them vulnerable under UNCLOS's anti-piracy rules.

No flag is rare because that immediately opens them to anti-piracy.

But coming back to my original point: it isn’t America’s determination that a registration is fraudulent. It is the flag state’s.

> it isn’t America’s determination that a registration is fraudulent. It is the flag state’s.

Sort of. If there is no flag, it's America's determination. And in many of the seizure cases, the flag state confirmed a fraudulent registration. (I believe there was one around Venezuela falsely registered with Panama.)

I’m not an AF vet so I don’t have an idea, but what’s the over-under that the US injuries were the crew trying to get the plane ready to fly after the alert came in? I think the number of injuries lines up closely with expected crew compliment.

Was hit on the taxiway. See story update: https://www.twz.com/air/images-purportedly-show-e-3-sentry-t...

Oh this is a good article. Thanks for that.

Towards the bottom they list some satellite imagery and a statement indicating they are possibly using the taxiways as parking.

Still leaves open the question of who might have been injured and where, but at least answers how the Iranians could have possibly hit a taxiing plane — they didn’t.

The Oklahoman and Jerusalem Post are also reporting it was a total write off.

Also adding: the spike in 2008 was transient and partially juiced by a weak dollar. Unfortunately, we will probably get no respite this time around.

At the current geopolitical trajectory, I also doubt $147 is anywhere near the limit of where oil is going.

I got what you were saying. I read it the same way. I’m sorry for your loss.

No one leaves this planet alive, and the best you can hope for is that the majority of your time is spent relatively healthy and independent.

I don’t disagree with any of your assessments, but I don’t know if it’s a bigger mistake than Iraq…yet. That war was a 10 year (longer if you bc point ISIS) debacle that cost trillions.

Let’s wait a few years before saying this mistake is bigger first.

However, one point that I agree with that might lead to this war being worse: the Gulf are showing some serious buyers remorse with sticking in the US orbit. Both the uselessness of America’s strategy and the almost clear prejudice Trump shows towards the Arabs vs Israel in the decision tree of this conflict is unsettling for the Gulf states.

Good write up. Not to mention the other big HR constraints on DoD engineers: they almost always have to be a “US person.”

Anyone who gets a CAC working on a personal computer deals with this all too much. The root certs DoD uses are not part of the public trusted sources that commonly come installed in browsers.

lol I very nearly included a rant about that but decided it was too far off topic. Not being able to smoke weed may be more of an obstacle these days though.

I haven’t touched a Palantir system since 2008 (and I still feel dirty) so I’m not the most read up on this: but Maven is just the harness or workflow tool. It still needs an LLM to evaluate data, and they used Anthropic for the kickoff of this war.

They made everyone more productive. But that also means everyone has to do more to keep up with said productivity.