leetharris

The challenge is that our bar for "smart" keeps going up and up.

If you compare to "smart" from a year ago, you have endless options that are basically free at this point which outperform everything from a year ago.

Deepseek Flash V4.1, Sonnet 5.5, GPT6 Luna, etc are really strong for the cost.

This is an incredible idea. I would usually be skeptical that such an engineering-focused tool would "make it," but Mitchell Hashimoto is one of the very few people I think can pull it off.

This is going to change so many things if it starts getting adopted.

I have the pro account for ChatGPT, Claude, Gemini, and Grok.

They all have various strengths and weaknesses. My favorite is still ChatGPT, then Gemini/Claude, then Grok.

Grok often feels 1-2 generations behind the competition in general use, but it has three things that I love:

1. It seems to be the best at understanding current events. Maybe due to X integration, or some other tool call optimization in the backend? I don't know, but I often ask about things going on, and the other models have outdated info, give unhelpful answers, etc.

2. It is generally the least sycophantic for personal things. Anthropic is getting here too. ChatGPT and Gemini are working on this, but previous models in those families would almost never say anything negative about what I am doing. Sometimes I need career advice, personal advice, etc and I like the tone of how it responds. I think Claude will be caught up soon.

3. For professional work, there are certain topics that other models would refuse to engage with. At my last company we had an enormous amount of legal users. When a deposition would need a summary on certain topics, most models would refuse. Grok would not. I understand the need for safety and I don't blame the other model providers, but for some professional use cases you NEED a model that is capable of handling sensitive subjects.

The only mistake he made was not buying Tesla around the M3 launch. Elon was desperate and would have sold it to him for cheap. He didn't take the meeting.

Otherwise I completely agree, once Tesla reached takeoff, it's too hard for anyone else to do it without burning mountains worth of cash.

I will say this in the most charitable terms I can. Microsoft simply does not have it in their culture to compete with something like this. Their prime days are over. They are slowly becoming IBM.

They were completely correct to not compete in foundation models. They would have no chance. I mean, they can't even make a decent app or harness to use the other models!

They aren't publicly traded.

None of them do it well from our experience. We had to write our own custom pipeline with a mixture of legacy CV approaches to handle this (AI contract analysis). We constantly benchmark every new multimodal and VLM model that comes out and are consistently disappointed.

"Polluting" is a very charged term. These satellites provide immense value. So far, there is no evidence these will stop us from watching the stars.

Document Crunch | REMOTE(US) / HYBRID

Company description: We do construction compliance AI. We're in series B, about $40m raised. Growing a lot! Feel free to reply or DM me if you want some more info.

Tech Stack: Typescript full stack except for Python for ML, NestJS backend, NextJS frontend, React Native mobile, IaC in Pulumi using TS.

Open Jobs:

- Senior/Staff/Principal Frontend Engineer (fullstack Typescript nice to have)

- Senior/Staff/Principal Node Engineer (fullstack Typescript nice to have)

- Senior/Staff/Principal QA Engineer

Location: Austin or Texas preferred, but open to remote for the right candidate in the US.

Jobs page: https://www.documentcrunch.com/careers#open-roles

For the roles not listed yet, email us directly at [email protected]

I used to work in ASR. Due to the nature of current multimodal architectures, it is unlikely we'll ever see accurate timestamps over a longer horizon. You're better off using encoder-decoder ASR architectures, then using traditional diarization using embedding clustering, then using a multimodal model to refine it, then use a forced alignment technique (maybe even something pre-NN) to get proper timestamps and reconciling it at the end.

These things are getting really good at just regular transcription (as long as you don't care about verbatimicity), but every additional dimension you add (timestamps, speaker assignment, etc) will make the others worse. These work much better as independent processes that then get reconciled and refined by a multimodal LLM.

It is insane that the thoughtful act of gift giving has been reduced into a checklist item. What is the point anymore?

Two PMs left a company with 125,665 employees. Yes, they are very important, but this happens every day at companies all around the world. Could be burnout, they are rich, got better offers, anything really.

People are so desperate for Tesla to fail they will latch onto anything.

> I wish .NET was more popular among startups, if only C# could get rid of the "enterpisey" stigma.

Too hard to ignore the benefits of cross-stack gains in Typescript/Python. The C# native phone, Blazor, etc just isn't quite there yet. Tried it at the last company, and full stack TS was just so much easier to do.

The reality is that the vast majority of startups don't make it. The #1 thing startups should be focusing on is hiring the right people and product velocity. TS just makes that easier in my experience.

Document Crunch | REMOTE(US) / HYBRID

Company description: We do construction compliance AI. We're in series B, about $40m raised. Growing a lot! Feel free to reply or DM me if you want some more info.

Tech Stack: Typescript full stack except for Python for ML, NestJS backend, NextJS frontend, React Native mobile, IaC in Pulumi using TS.

Open Jobs:

- Senior Frontend Engineer (fullstack Typescript nice to have)

- Staff Analytics Engineer

- Senior Product Designer

Location: Austin or Texas preferred, but open to remote for the right candidate in the US.

Jobs page: https://www.documentcrunch.com/careers#open-roles

This is kind of a wild story. Sam Altman is openly flexing that he was able to skirt the rules and regulations by threatening economic damage to California. It's not even subtle anymore.

This reminds me of when the former CEO of Hyundai, Chung Mong-koo, went to prison for embezzlement. In just 3 years he was pardoned because the President of South Korea basically said, "we need you for the economy."

We're not even pretending that the government is in control anymore. It's just full on anarcho-capitalism on display.

Incredible technology, but that was an insufferable video. Still very cool, I might preorder one!

I have the Xbox Ally, Steam Deck, and original Ally.

The original Ally software launch was a disaster. Unbelievable amount of bugs and overall terrible user experience. After 6+ months of updates it was decent.

I figured, hey, maybe they figured it out in advance this time? So I pre-ordered an Xbox Ally.

It is a complete disaster in terms of software. It took 90 minutes to setup and download initial updates on a Google Fiber connection. Things break constantly.

The other day, I got a new error, "Something went wrong and your PIN isn't available." When I try to click anything, it just goes black. After 6 or 7 restarts, it randomly glitches out and takes me right to desktop without any PIN.

It is just constant bullshit like this. The entire experience breaks over, and over, and over. I hate it so much. Back to Steam Deck.

> there was a time when either Boeing or the military industrial complex would handle these things with precision and delivered ahead of time.

Is this a joke? Boeing or similar delivering ahead of time?

People are usually not querying across 5 million documents in a single scope.

If you want something as simple as "suggest similar tweets" or something across millions of things then embeddings still work.

But if you want something like "compare the documents across these three projects" then you would use full text metadata extraction. Keywords, summaries, table of contents, etc to determine data about each document and each chunk.

Full text agentic retrieval. Instead of cosine similarity on vectors, parsing metadata through an agentic loop.

To give a real world example, the way Claude Code works versus how Cursor's embedded database works.

Embedding based RAG will always just be OK at best. It is useful for little parts of a chain or tech demos, but in real life use it will always falter.

For anyone wondering how to quickly get your codebase into a good "Gemini" format, check out repomix. Very cool tool and unbelievably easy to get started with. Just type `npx repomix` and it'll go.

Also, use Google AI Studio, not the regular Gemini plan for the best results. You'll have more control over results.

This is the nature of public markets. Not everything should be public. In fact, MOST things should not be public. Because being public forces accountability and liability to shareholders in a way that is completely unlike being private.

A company can be successful by most metrics, but if certain trends are not heading in the right direction, then faith in the stock drops, employee compensation goes down, future investments become dicey, etc.

This is the nature of public companies. This is why they don't want to be public anymore.

Only for Sonnet. No 1m for Haiku (this new model) and Opus.

This means 2.5 Flash or Grok 4 fast takes all the low end business for large context needs.

The main thing holding these Anthropic models back is context size. Yes, quality deteriorates over a large context window, but for some applications, that is fine. My company is using grok4-fast, the Gemini family, and GPT4.1 exclusively at this point for a lot of operations just due to the huge 1m+ context.

Ok so Musk is the reason mass transit isn't a success?

So why didn't we get mass transit before Musk was rich?

In Austin we approved an enormous rail project and the taxpayers approved billions to build it. Years later, they have descoped it by 75% and they haven't even come close to delivering that small bit they reduced it down to.

Was that Musk too? Or is overregulation destroying our ability to innovate?

Ah yes, the hyperloop is what killed California high speed rail. If only it weren't for those few tweets, it would be successful and done by now.

Or maybe there's another reason these high speed rail projects consistently fail. An insane regulatory and litigious environment where no technology progress can be made. Meanwhile, in Asia, rail is being laid at insane pace.

What do you think it'll take to match their progress? Do you honestly think Elon is the reason this is all failing?

In the last 50 years, well before Elon was involved and supposedly sabotaging all this great subway progress, how many new subway projects in America were started and successfully finished?

Maybe Elon's solution is a requirement to make ANY progress because the other ones have stalled out?

I would love a subway. Here in Austin, they proposed a bunch of new trains and subways for billions of dollars. We all voted yes and funded it. After a multi-year study, they revised it to about 1/10th the scope and doubled the cost. It still hasn't materialized.

Do you all realize the problem here?

A breakthrough in tunnel cost and speed per meter. Have you read into this at all or is this just another snarky comment?

Can you list the significant harm to people and the environment? I read the article and I'm struggling to find anything.