BluSyn

Compared to what? All the comments seem to agree, but I curious if people here have actually used Grok.

I rotate between major models frequently. Grok has been up there in accuracy and research for some time, trading places with Gemini IMO. Latest 4.3 release has been solid.

Composer is pretty good and now they own Cursor. Don’t count them out yet.

So.. it’s bad, compared to what? Claude from 2 months ago?

I still stick with btrfs for this reason

Rust on embedded is fun to play with.

Few years ago I made a custom RGB LED rope light controller using ESP32 C3 DevKit and Rust embedded, connecting light to homeassistant via MQTT auto discovery. Was surprisingly easy to get started as someone who had limited experience with embedded programming, and only hobby Rust experience.

But the supported hardware in Rust crates was limited at the time and C3 dev kits weren’t widely available, so I never used it for anything. HAL support has only gotten better since then.

I may try to resurrect that project now with Pico 2 and Matter.

Strike on Venezuela?

Perhaps related? My main fiber WAN went out few hrs ago, failing over to Starlink backup. Discovered it’s a cloudflare issue, as my multi-wan setup tests against 1.1.1.1, which suddenly stopped responding (but only from my fiber ISP). Switched to testing 8.8.8.8 to restore.

If it weren’t for recent cloudflare outages, never would have considered this was the problem.

Even until I saw this, I assumed it was an ISP issue, since Starlink still worked using 1.1.1.1. Now I’m thinking it’s a cloudflare routing problem?

Steam machine so close to perfect, but 1x USBC and 1GB Ethernet are huge misses for a 2026 device. Also needs more VRAM. May be better to just do custom SFF build.

Unsecured government loan to a successful company to fund acceleration and growth is a "handout".

A "bailout" is what happened in 2009, in the sense the banks would literally have collapsed without it (and they probably should have).

OpenAI is not going to collapse without these loans. Huge difference.

Also for the record, not rationalizing, because I'm not in favor of either handouts or bailouts.

I'm confused about language, as "loans" to me do not equal "bailout". The equating of the two seems odd, as many government incentives use loans that pay back with high interest, so governments MAKE money on those kinds of deals.

Also clear that the 1.4T figure includes some accounting for spend that does not come directly from OpenAI (grid/power/data infra for example). Obviously some government involvement is needed, but more at EPA/State/Local level to fast track construction permits, more-so than financial help from Treasury.

I'm confused why this generates such sensational headlines.

grok-code-fast-1 is my current pick, found accuracy and speed better than Sonnet 4.5 for day-to-day usage.

I would expect AI to influence slang in future generations. Would be more surprising if it didn't.

Wasn’t sure if a random TestFlight link would be safe/wise to share, so shared original source.

FYI on X there is a TestFlight link to try it: https://x.com/jack/status/1941989435962212728

Surprised to see Jack pushing code himself. Love to see it.

How can Tesla advertise a “more accurate” number if they are required by regulation to use EPA estimate?

EPA range estimates being inaccurate is a real problem. They do not, and are not designed to, give actual expected range. It’s meant to be an “average” of “mixed” driving.

Take latest Model Y as example. If you compare EPA range vs WLTP (commonly used in EU)

327mi EPA est. (526km) US version (long range) / 586km WLTP est. (364mi) EU version (long range)

The WLTP is “average” as well, so which of these is more accurate?

This problem is not unique to Teslas, and actually not unique to EVs either. It’s just more noticeable, as ICE vehicles usually advertise MPG and tank size, not total range. So EVs suffer from their own advertisement highlighting numbers that will never be accurate.

We clearly have very different definitions of corruption. Actual aeronautics engineers with hands-on expertise offering advice on how to modernize air safety systems does not fall within my definition of corruption.

Noticed this also. It doesn’t feel organic.

You seem to be assuming that the full cost of the cluster is recouped by Grok 3. The real value will be in grok 5, 6, etc…

xAI also announced a few days ago they are starting an internal video game studio. How long before AI companies take over Hollywood and Disney? The value available to be captured is massive.

The cluster they’ve built is impressive compared to the competition, and grok 3 barely scratches what it’s capable of.

1- The conspiracy is that blocking signal links is some top down dictation from Elon as part of geopolitical power play over censoring dissenting opinions about DOGE. This is some next level mental gymnastics. There may be good reasons to temporarily block links to a specific service, eg, maybe an ongoing phishing scam, or security issue. Plenty of rational reasons that have nothing to do with Elon and Trump.

2- X is full of people arguing about DOGE, positive and negative, and Elon is constantly attacked on his own platform by both small and large accounts. These posts are not censored.

3- Elon routinely talks about the importance of free speech, and yet I keep reading claims he’s against it from people advertising bsky and mastodon, which absolutely do not (by any reasonable definition) represent free speech platforms. They are more heavily moderated than even old Twitter was.

4- I have seen no evidence of Elon having a “propensity for authoritarianism and censorship”. Every time he gives a speech he specifically talks about being pro 1st amendment, regularly responds to his detractors, and actively defends the constitution.

I do not know what version of Elon you see, but from my perspective you are talking about a cartoonish media caricature that is the opposite of the reality that I see. We appear to be see the same person, same events, and are drawing opposite conclusions. Hence my point about sense making no longer seems possible, even when we actually appear to agree. (If anybody comes out as pro-censorship, I would be the first to call it out.)

Do all these commenters believe this nonsense? The conspiracy thinking has reached new peaks. Hard to take concerns of free speech seriously from people promoting platforms that routinely “moderate” any dissenting opinions (HN included).

Our ability for collective sense making seems to be permanently destroyed. Two people who seem to agree on everything point-by-point then reach different conclusions on the final step. Its bewildering.

Did these digital modernization efforts result in lower agency head count, or did they instead use it as excuse to ask for higher budgets?

Real modernization of systems would result in efficiencies that allow for reduction in staff. Tech exists to be a lever on productivity. This almost never happens; historically bureaucratic modernization efforts only proceed if absolutely necessary for core functions, and strangely seem to result in higher head counts. IRS is a great example of this. Their headcount has doubled while they supposedly "modernized". Digital systems should increase productivity of the average IRS employee, instead they just invent more busy work.

This feels like hyperbole. Nobody is arguing to destroy the village, the federal government is not going to disappear due to 20% spending reduction. Reducing spending from ~$6T/year to ~$4T/year is not the end of government, which is the explicitly stated goal of DOGE.

Calling it a cyberattack is hyperbole. Lost objectivity points from the outset.

The split of opinion on this in tech circles is quite surprising.

No agency would voluntarily modernize systems, which would inevitably reduce head count and put half of them out of jobs. This has been a on-going fight for 30 years. Every politician who previously tried to modernize agencies failed due to intense internal resistance.

Everyone in tech clearly fell on the modernization side. This is what many of us wanted since the 90s. Finally happening. In real time. Headed by one of the greatest tech disrupters since probably Edison. Now they act like the sky is falling?

I can’t find any steel-man argument against DOGE. No modernization plan proposed the traditional way through committees, consultants, contractors, has worked or will EVER work. You have to rip off the bandaid the hard way.

Only explanation I have for those opposing this is some combination of personality derangement spread by nefarious interests, financial incentive, or some crazy model of the world that glorifies bureaucratic power as some fundamental right enshrined in constitution.

Imagine being on the side of the Empire and trying to stop the rebels as they infiltrate the Death Star.

I violently recoil at comments like this.

Elon is so effective as a leader he seems to break people’s brains. No other person could have started this company and had even half this success. There’s a reason all the most talented flock to his companies, despite “conventional wisdom” saying they shouldn’t. It takes a lot of self deception to ignore the reality that he obviously must be doing something right.

Reasonable take not allowed. Must be for or against so I know what ideological bucket to place you in based on current tribal alliances that may or may not change.

Using AI for market price discovery actually sounds like a valuable use case.

My experience with Prisma, which has a very similar DSL for defining schemas, has changed my mind on this. Makes me much more productive when maintaining large schemas. I can make a one line change in the schema file and instantly have types, models, and up/down migrations generated and applied, and can be guaranteed correct. No issues with schema drift between different environments or type differences in my code vs db.

Prisma is popular enough it also has LSP and syntax highlighting widely available. For simple DSL this is actually very easy build. Excited to have something similar in Rust ecosystem.

The entire plan hinges on low cost high volume. That means reducing parts and complexity. Stainless steel body so no paint shop or variations to worry about, or paint maintenance, etc. removing steering wheel, removing glass roof, removing doors, and more.

For high density you have the robovan.

They literally showed a clip of a robot cleaning the car in the announcement. Also perfect task for Optimus.

Just in California, right? What’s the prevent a studio elsewhere from doing this? Online distribution makes the legal borders meaningless here. So people in California will just need a VPN to watch future action movies?