subcan

Just a guy that likes to do IaC

[subcan @subcan1]

@MikeWingerii I deleted youversion about 5 years ago. My daily driver is The Blue Letter Bible. Although BLB is way better on android. still better than any of the others on iPhone.

[subcan @subcan1]

so I bought a subscription of grok heavy and ran out of use in about 30 minutes of chat. I don't know how that could be usable for coding? What a frustrating day. Then grok argued with me about not being able to solve a problem that chatgpt 5.5 solved in two tries... frustrating day. @elonmusk @grok

[subcan @subcan1]

@NathanMcNulty it is Microsoft and they use copilot which sucks... haha

[subcan @subcan1]

What horrible advice... if you don't disengage when FSD is doing stupid stuff then how will @Tesla_AI get good training data. Every single time FSD "figures" it out for you... FSD did not learn and will figure it out again next time.. that is how AI works.
What you said is advice because people listen to you.

[subcan @subcan1]

yeah, and my 25 M3 no longer backs into my driveway which is needed to charge. And then the new "what happened" when you disengage is four choices that does not cover that case... so now I am calling it a critical disengagement...
I like the drive, but now FSD is doing so many dumb things that it did well before... @Tesla_AI @elonmusk

[subcan @subcan1]

@Tesla_AI @elonmusk There really needs to be a way to add a grok system prompt with tesla app. I have to express my rules at the start of every conversation and each time it is a little different. Also takes away from the reason I am starting chat with grok and then I need to collect my thoughts.
A system prompt is critical, especially with the cars grok, which does not default to professional conversation and prefers slang and jokes, which is not interesting to me when dealing with an LLM

[subcan @subcan1]

@XFreeze here is chatgpt and grok with same answer, whil gemini gives the best answer and Claude gives the "counting 10 numbers starting at 11" answer.
It is all about context and what is meant by a trick question.
Not sure what you think you have "proven" by your biased post

[subcan @subcan1]

@TaraBull give it a try. My guess is that this guy has loaded a misleading system prompt.

[subcan @subcan1]

@tesla_na A week ago a mobile technician came to rotate the tires on my 2025 M3 performance, but could not finish the job since his car jacks couldn't fit under the car? So now I am told that I have to go to service center for simple tire rotation? hmmm

[subcan @subcan1]

Some people are totally amazing

[Dweller @One_Way_Home]

This is an under appreciated art in itself - surgical precision at it’s finest.

[subcan @subcan1]

@BLKMDL3 my biggest hope is that FSD stops the crazy right lane bias. I want left lane bias. Almost always the best lane to be in. The inner left turn lane, or left lane on city roads. Only right lane when a right turn is coming up.

[subcan @subcan1]

I value truth above all else. The distinction between hallucinating and bad advice is subtle. I agree that Grok seems to have few hallucinations, but I have also noticed, since the last version, that the answers for complex reasoning have been subpar. Almost like expert mode is taking shortcuts

[subcan @subcan1]

@elonmusk my 2025 M3 "Chicken Little" ran over a pool squirrel a few days ago. To be fair it felt like the squirrel tried to get run over, but Chicken Little, who is scared to change lanes, was not scared to run over a defenseless little squirrel

[subcan @subcan1]

I asked grok 4.2 beta the exact question and here is the answer
"**Result**
Cannot be answered yes or no. Whether President Trump was “right” to authorize U.S. strikes on Iran is a normative question without an objective factual resolution.

**Facts**
- On 28 February 2026 the United States and Israel began large-scale strikes on Iran under Operation Epic Fury, targeting nuclear sites, missile infrastructure, naval assets and leadership targets.
- Iranian Supreme Leader Ayatollah Ali Khamenei was killed in the operation; Trump stated that much of Iran’s senior leadership was also eliminated.
- Trump’s stated objectives: prevent Iran from acquiring nuclear weapons, destroy its missile and naval capabilities, and enable regime change. He cited Iran’s rejection of prior negotiation offers (including after June 2025 U.S. strikes on three nuclear sites) and the need to avenge three U.S. service members killed in Iranian retaliation.
- Iran has launched retaliatory attacks on Israel and U.S. bases in the region.
- As of 1 March 2026, Trump has stated that combat operations continue in full force until all objectives are met, that additional U.S. casualties are likely, and that the operation is “ahead of schedule.”
- Public reporting contains no indication of prior congressional authorization for the 2026 operation.
- Earlier actions include the January 2020 Soleimani strike (U.S.-justified as imminent-threat self-defense) and June 2025 strikes on Iranian nuclear facilities.

**Assumptions**
- “Right” could refer to legal authority under U.S. or international law, strategic prudence, or moral justification.
- Media reporting on casualty figures, leadership deaths, and exact damage is based on statements from U.S., Israeli, and Iranian sources and remains subject to verification in an active conflict.

**Inference**
Under one view, the strikes are lawful and necessary self-defense against a nuclear threshold state that has attacked U.S. forces and proxies for decades. Under another view, the operation exceeds presidential authority without congressional approval and violates the U.N. Charter prohibition on the use of force absent an armed attack or Security Council authorization. Strategic outcomes (nuclear setback versus risk of wider war or regional destabilization) cannot yet be measured.

**Limitations**
The conflict is ongoing (1 March 2026); casualty totals, long-term effects on Iran’s nuclear program, and final strategic results are unknown. Normative judgments of “rightness” depend on the observer’s weighting of security, legality, and proportionality and are not resolvable by evidence alone.

**Confidence**
Certain on the occurrence and stated justifications of the reported strikes.
Uncertain on ultimate legality, morality, or net strategic benefit."

[subcan @subcan1]

@thekitze yup that is the story of the vibe coder.. haha

[subcan @subcan1]

@aigkenham @elonmusk this video was masked as adult only.. how crazy. I thought X was free? there is a serious problem with grok 4.2.

[subcan @subcan1]

@beholdisrael another amazing concept here is Khamenei was a tyrant and scourge to the earth for 47 years... hmmm 47... hmmm

[subcan @subcan1]

@elonmusk @Tesla My FSD does not experience rain often so I don't think about the challenges. How about compressed air vents at cameras? One compressor could handle supply. The air would form a shield-ish from rain/snow even touching glass, and any debris would be blown off. Doesn't seem like it would take high pressure air. Would be easy to route lines to all vents, including rear camera. Side pillars could have air from above the camera. Just a thought

[subcan @subcan1]

@ns123abc AI is a high stakes game. Is anthropic for America or not. If not then they should not reap the benefits of riding the edge.

[subcan @subcan1]

@Tesla_AI @Tesla @elonmusk Today I renamed my 2025 model 3 performance "Chicken Little". I will name it back when FSD starts acting with an ounce of courage.
FSD used to actually take the small lane opening that someone gave. Now it needs hundreds of clear feet, and then someone behind me just takes the spot. It is painfully crazy.
What I wonder is, was the car doing safe lane changes in the past or was it just "taking chances?"
FSD is awesome, however Chicken Little needs to be trained with some new data...