Hacker Newsnew | past | comments | ask | show | jobs | submit | e9's commentslogin

Interesting, I got notification from Github:

"Incident with Grok 4.6 Copilot AI Model Provider"

at 7:20 am PT


Someone could've trained model on top of GLM. Same way Cognition trained their SWE model on top of Kimi and Cursor did same with their Composer model.


While possible the amount of variation in serving infrastructure is unlikely to land with actually giving the exact same errors zhipu does.

It feels like glm flash, and there was a report zhipu had secured a huge new cluster suggesting they have the capacity. My guess anyway.

https://www.tomshardware.com/tech-industry/artificial-intell...


The reasoning levels are the same as GLM 5.3. GLM 5.3 is still not open...

I believe it's GLM 5.3 Flash or Air.


Reasoning levels are often just injected system prompts so not a great way to fingerprint models.


But it's an error, not a response.


I agree with you on the harness. I find that Claude can be good in any harness but GPT is only superior inside Codex.


Antares were the first ones: https://www.energy.gov/articles/department-energy-celebrates...

And closely followed by Valar Atomics two weeks later: https://www.energy.gov/articles/department-energy-celebrates...


Actually this guy achieved criticality before either of them: https://i.imgflip.com/a0w01q.png


I want to propose alternative reality where 1.5-2.5T in value doesn't go to a handful of companies. Instead it turns out to be like restaurants where this gets distributed to lots and lots of small, local, mostly interchangeable teams. There will of course be some super star "chefs" leading the industry and setting trends and some "restaurant chain" like big businesses and supply chain for all of this.


FWIW I do think that availability of competitive open weight and other non-frontier models, along with improvements in harnesses that can get good results out of these models, will result in less concentration and a healthier marketplace.

However, these frontier labs are also making moves that could let them capture a disproportionate share of the upside. One possibility is a situation analogous to the smartphone manufacturing space, where there are dozens of players but just a handful (e.g. Apple, Samsung in smartphones) capture the lion's share of the revenue.


Apple you can’t exit the ecosystem.

Samsung the same. And is the best android device.

If tomorrow comes a Nokia os will be dead in the water: it has no apps.

But with a new llm that doesn’t matter. There is nothing sticky about typing Gemini, Claude or codex in a cli.


There's nothing sticky today but you can bet they're working maniacally to fix that. These companies will make most of their money in the enterprise space and there are probably unlimited ways to engineer stickiness in an enterprise setting. Like, MSFT still rakes in those billions despite pretty much every one of their products having commodity competitors.

The AI labs are also making moves to secure long-term enterprise presence, such as their Forward Deployed Engineer strategy. I think that is a trojan horse play that could make enterprises dependent on them forever, much like so many companies are still dependent on IBM's mainframes. As an extreme example, you could imagine a company's core business logic encoded in the weights of a proprietary model custom-trained and hosted by one of these model providers, something even more inscrutable and sticky than ancient COBOL codebases.


The world is not zero sum. Value is created, not just preserved. Anthropic and OpenAI creating value does not imply that smaller guys can not also create value.


But marketplaces also exist and big players in a marketplace are often able to manipulate the market such that they are advantaged and small players are not able to break in.


This is true of every market that has ever existed, and that's not stopped small players from finding niches.


How? Training and operating models seems to naturally focus on those willing to invest quite significantly in these operations.


If RAM prices come down, running your own models will be relatively affordable.


Sysco is pretty big.


Probably because they turn off transponders while crossing that area.

"Traffic is generally picking up in the strait, with several laden tankers seen exiting into the Gulf of Oman over the weekend, though some of them turned off their transponders. The latest was a Greek-flagged tanker carrying Iraqi crude to Singapore."

https://financialpost.com/pmn/business-pmn/iranian-crude-oil...


Unfortunately this author has no idea about how equity works in real life, 2 things that stand out from experience of being very early employee 8 times: 1. If startup raised money at some point in the past and they are at 50M valuation then there is no chance average employee will get offer of 1% (you would need to be someone special to get that) 2. There will be many more dilution rounds before they get to 1B valuation so chances are that 1% will get cut to 0.3% over that time period


You give two critiques relating to the exact numbers the author has choose (equity package at join, valuation at exit), neither of which is really related to the authors hypothesis that people misvalue EV from equity.

You can slide these numbers around however you want and the point still stands, even if you disagree with it for other reasons


my 2nd point says that due to dilution rounds your 5% chance is not 10M but about 3M so

5% * $3M = $150k expected equity value

$150k / 1.15 = ~$130k/year

which is same as you started with


you can get 1% as a founding eng at seed, and its not uncommon for a 5 at 50 seed

dilution is also dependent on the opex, founder negotiating power, and growth of the company. There are startups raising monster rounds at <5% dilution a round

If you are an employee however and your co is raising highly diluted rounds with poor growth probably best to jump ship


I recently worked with NRC dataset, specifically about nuclear reactor events and status reports(example: https://www.nrc.gov/reading-rm/doc-collections/event-status/...). Public data that just needed some cleaning. Several time Claude API would refuse to engage. Because of that I can't trust Claude to clean production data sets.



Shit, that was 10 years ago already? Feels more recent, also in terms of technology.


Private companies are not obligated to report GAAP numbers which means they can creatively twist numbers while still be truthful until they go public


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: