It's kind of funny how not a lot of people realize this.
On one hand this is a feature: you're able to "multishot prompt" an LLM into providing the wanted response. Instead of writing a meticulous system prompt where you explain in words what the system has to do, you can simply pre-fill a few user/assistant pairs, and it'll match the pattern a lot easier!
I always thought Gemini Pro was very good at this. When I wanted a model to "do by example", I mostly used Gemini Pro.
And that is ALSO Gemini's weakness! Because as soon as something goes wrong in Gemini-CLI, it'll repeat the same mistake over and over again.
> despite things being, well, entirely underwhelming
People use LLMs for all kinds of things, but for coding it is absolutely not underwhelming. Can you treat it as a real independent developer, one that doesn't need supervision? No. Can it save you hours and hours of work? Yes.
I switched to Ly a few weeks ago, after SDDM kept crashing. And I switched to SDDM because GDM did the same thing earlier. They both were caught in some kind of crash-loop that it was impossible for me to switch to another virtual console.
I think you're both correct. Gemini is _still_ not that good at agentic tool usage. Gemini 3 has gotten A LOT better, but it still can do some insane stupid stuff like 2.5
I noticed one single API error a few hours ago. Didn't seem to be down for long.
(I prefer the occasional downtime here and there versus Gemini's ridiculous usage limits)
The furthest distance a robot on mars has traveled from its landing position isn't even 50km. Over many, many years.
For example, on Nasa's website it says this about Perseverance:
> This map view shows the route NASA’s Perseverance Mars rover has taken since its February 2021 landing at Jezero Crater to July 2024, when it took its “Cheyava Falls” sample. As of October 2024, the rover has driven over 30 kilometers (18.65 miles), and has collected 24 samples of rock and regolith as well as one air sample.
That's about 8.5km per year.
So I think they would have to land a new one pretty close.
Unfortunately, the roadway infrastructure on Mars hasn’t been updated in decades, so there are lots of rocks and potholes/craters and other obstacles that need to be avoided. You know how those jeeps go out to Moab, Utah and do the 4x4 trails? It’s like that but there’s only one vehicle and no human and 1 hour communication round-trip and if something goes wrong the tow truck is millions of miles and billions of dollars away
The two solar powered ones are decommissioned, both mostly because of diminishing power but one got stuck first. Away from the equator solar power is less effective right? And it's colder, which exacerbates a serious problem: most the energy goes to heating the main components.
The nuclear rovers are doing their assigned missions, and can go about 100m/day IIRC. So, 5000km trek to the pole would take about 50,000 days at 10d/km. Give or take a few thousand km. (It's a whole planet right?)
This is all Wikipedia level research and from memory.
One major issue with transportation in the Martian environment is the extremely abrasive dust and the sharp rocks. Pretty much every rover has had the issue that the wheels deteriorate very quickly and dust gets into every nook and cranny, eventuelly destroying important movement-related mechanisms. As to their movement speed, that's mostly down to the movement being manually commanded and with the light delay of about 20 mins (one-way), you can only command the rover to go so far before involuntarily hitting an object.
I recall reading that a major candidate for any early colony is in lava tubes, dust on the would be one factor, but radiation shielding is another. Either you have to ship materials from Earth and build them, consume whatever is available and useful locally, or make use of whatever Mars-nature provides. If you can get away with lighter materials to build below surface then it seems better compared to more durability/shielding requirements above.
Exactly. There is no wind. All the little solidified impact glass particles, with their razor sharp microscopic edges, have not been smoothed by even the slightest wind erosion.
Last night, just after Gemini 3 was released and became available for Gemini-CLI, I saw Gemini-CLI's team post that you could access Gemini 3 with either an API key OR with _Gemini AI Ultra_, so I thought: great, I'll get that!
Now you CAN NOT get the Google One stuff if your account is part of a workspace.
I thought: how awful. I want to pay, but I simply can't?
Oh, but then I noticed: You CAN add a _Gemini AI Ultra_ license via the Google Workspace Admin area, great!
Turns out: you fucking can't. That's _Google AI Ultra FOR BUSINESS_ and that IS NOT supported.
So I had to get the Google One subscription on my personal account after all.
Combine that with the _pathetic_ usage limits: somehow not token-based, but amount of requests per 24 hour window (which is 500 for Gemini 3) and Gemini 3's incredible chattiness (it uses A LOT more requests to get something done compared to Claude) and you hit the usage limits in just 2 hours.
On one hand this is a feature: you're able to "multishot prompt" an LLM into providing the wanted response. Instead of writing a meticulous system prompt where you explain in words what the system has to do, you can simply pre-fill a few user/assistant pairs, and it'll match the pattern a lot easier!
I always thought Gemini Pro was very good at this. When I wanted a model to "do by example", I mostly used Gemini Pro.
And that is ALSO Gemini's weakness! Because as soon as something goes wrong in Gemini-CLI, it'll repeat the same mistake over and over again.
reply