My primary use is AI coding agent. Its vastly cheaper than Kimi K3 and I haven't found a scenario where I really need Kimi K3 versus smaller models. GLM-5.3 Flash is good but there is series of bugs in the vllm middleware that prevent GLM models from getting all of their reasoning content returned to them that impairs inference quality. A lot of inference providers use vllm which makes it hard to find a good provider for GLM. I've been using friendli.ai but using GLM-5.3 Flash from them is more expensive then using DS V4 Flash from deepinfra.com simply because deepinfra.com is so cheap. The DS V4 Flash cost at together.ai is similar to the GLM-5.3 Flash from friendli.ai or at least that's what I found in my benchmarks a week ago: https://www.linkedin.com/posts/joshheitzman_i-ran-a-fuller-r...
4.1 consistently surprises me in capability for the price. And I don't think I'm the only one. It's been dominating the leaderboard at OpenRouter, and I just got an email today from Fireworks saying they were _raising_ the price by about 30%. I'll probably switch, because their infra doesn't support being the highest-cost, but it's still telling.
I tried it a few times and liked the speed, but often found it ended up looping, i.e. repeating the same token sequence (e.g. the same sequence of 5 paragraphs) over and over again until it hit the max output limit. This doesn't end up happening every session, but does every now and then.
My impression of DSv4.1-flash was very positive aside from this. But that was enough for me to stick with GLM-5.3(-flash), which both gave me consistently great results
I was using a vibe coded bare bones harness.
I was wondering if this was normal from DSv4.1-flash, or if its my harnesses fault.
Spot on. Py2 was integer division, Py3 was true division. In Py2 you could get the true division with an `from __future__ import division` (Can't recall which Py2 version this became an import). Fun fact too: In Py3 you can do a // to get back to the floor division effect (e.g. 1 // 2 --> 0)
PEP 238 lays out the story if anyone is curious. Well into the Python 3 transition, I still had a habit of multiplying the denominator by 1.0 to ensure it was a float.
It's going to continue to suck getting prompt rejections on their overly aggressive "benign" safety margin thresholds [1]. Without a clear understanding of what triggers it, I guess we will pay for input/output tokens on the usage credits and not get refunds? As others on HN, I found very limited utility during the Fable 5 week in early June for my benign-engineering tasks and don't look forward to this new release, but I will try.
Would you consider .aero, .cat, .asia, .info novelties? They have been around for 20+ years. Sure, there are over 1500 gTLDs now, but when does something stop becoming a novelty? .ai a ccTLD, that Google recognizes as a gTLD, is that a novelty? These are a bit rhetorical.
I can't speak for the others, but registering a .aero domain requires submitting an application that's reviewed by a human and assessed based on the domain's relevancy to the aviation industry.
I would define them by the bar to entry. When a spammer can pick up thousands of domains for a few hundred bucks or less because a dodgy registrar is intentionally selling all domains on a specific TLD for next to nothing I find it safe to block anything on that TLD. I have always done this for email both personally and professionally and never once received any push-back. I explained the risk of collateral damage and leadership was always fine with it. Customers were B2B and could log in to open a ticket. Customer emails would come from domains we already trusted in their on-boarding process.
Has either of these companies released models before this? It's hard to believe that they could release a supposed Mythos-level model just out-of-the-blue. Deepseek, Z.ai, Alibaba/Qwen have been at this for a lot longer and have been releasing models with steadily increasing capabilities for about 18 months now. I find it hard to believe that these new companies would just suddenly release a Mythos-level model without releasing anything prior.
But did they deliver anything yet?
There are many startups with notable founders that are not being able to get models launched. And even if they are able to launch, so they get PMF.
In case of Sakana, they clearly focus on the Japanese market an have buildup a good pipeline on sovereign AI. But similar to Aleph Alpha or to a certain extend Mistral, I don't see how they can keep up.
I live in Japan and yet can't seem to pay for their API in JPY... I bet their enterprise customers don't have that problem but it was pretty annoying given "AI in Japan" appears to be there only selling point.
The only way they can compete in Japan is the enterprise-game. Their partnership with Daiwa and MUFG is probably exactly what they should be doing. I doubt that they get that far though. Like Mistral who has partnerships with BNP Paribas and Airbus, they deploy in an on-premise or private cloud and in those settings, their models make good PoC's but cant compete where it gets interesting -- Workspace Agents. If you look at them from a Chatbot over RAG perspective, maybe they can do it. But that's tech from 2023.
I think every country/major country at the government level will back any home grown talent in view of USA restrictions.
The models don't need to be as good, just good enough for the task. I am using a Claude Q3 2024 model mostly (Haiku 4.5) at present, and it delivers what I need.
How do we define "mythos level" exactly outside of marketing buzz? I don't even think the majority of us can access Mythos yet even to make a comparison.
Mythos is extreme hype. We are at a combo of authoritarian politicians peddling fear for power and tech bros trying to extract maximum investment returns.
We’re going to have many LLM/toolset combos that do what mythos does.
I have so many stories about his absolutely terrible behavior at conferences. He once refused to pay the entry fee to a charity event and had to be physically ejectedy.
Absolutely better at PR than any actual work, pay careful attention and none of his early stuff was particularly novel, from a technical perspective.
But for whatever reason, we venerate him just because he was victimized by the state. The world is not a dichotomy -- sometimes bad things happen to bad people.
He got all of the "Free Kevin" attention because of how long he was left in jail before trial and then being stuck in solitary confinement after sentencing for months.
If he had been treated fairly by the justice system he wouldn't have gotten nearly as much attention.
He was also autistic, a lot of the behavior can be explained through that lens.
>He got all of the "Free Kevin" attention because of how long he was left in jail before trial and then being stuck in solitary confinement after sentencing for months.
That was uncalled for on the part of DOJ.
>He was also autistic, a lot of the behavior can be explained through that lens.
I'm autistic. Maybe I should go commit a bunch of felonies to increase my chances of a good job and stature in the hacker community, since things like publishing code, publishing peer reviewed papers, and mentoring newbies have not been productive ways of finding gainful employment nor respect of my peers.
I have friends who did things like take a gap year to travel the world or met their spouses on nights I stayed in to study, and some evenings when browsing HN I feel very sad that I wasted my 20s on a society that does not care about me.
Anyways, sorry to wall of text, but what you said really struck a nerve with me -- there are hierarchies in any community, and one thing I've noticed with the hacker scene is one group of people can mess up over and over using the same sets of facts or diagnoses, but others can expect to have worse outcomes with better behavior for reasons that elude me to this day.
> I have friends who did things like take a gap year to travel the world or met their spouses on nights I stayed in to study, and some evenings when browsing HN I feel very sad that I wasted my 20s on a society that does not care about me.
I'm glad you have finally recognized the problem.
Stop living for your idea of others and start living for yourself.
>Stop living for your idea of others and start living for yourself.
Unfortunately, I exhausted my emergency fund trying for certs and attempting to escape a domestic violence situation, so I do not have that option. I'm trapped.
Kevin was famous for being mistreated by the DoJ and writing some books which were perhaps not particularly true in hindsight. After he got out of jail and rejoined the community he lost a lot of respect for being himself, though it's not impossible that years of imprisonment and a long time in solitary had some permanent negative effects. In other words... you shouldn't envy Kevin's life.
For the rest: nothing's stopping you from having fun, regardless of age.
Maybe, but if I am looking for a female spouse, a mens conclave is probably not the best place to find one. (I would assume the audience there is largely male?)
But well, he also is looking for respect and regocnition among his peers and vienna is a nice city.
> You should be safe in Australia since they actually need to be friends with China.
As a non-Australian, I enjoyed very much reading about Australia in a book by Tim Marshall titled "The Power of Geography". I didn't quite realize just how vulnerable they become with China's ambitions and expanse in the South Pacific due to their reliance on vital sea lanes for trade with its Asian partners. After reading that book, I can appreciate your comment much more.
reply