This is a temporary problem. A 10x-100x speedup will allow us to simply wait for the response, instead of finding something else to do in the meantime.
I wonder what the experience is of people here with the ultrafast mode (8x) of Codex.
I've tried to find a combination of a good-enough model, a VERY fast provider, and cheap enough serving to make this all doable. DeepSeek itself is quite fast, but I crave even more speed (and ideally something hosted outside of China).
Fireworks and Baseten seem to be the leading providers that are available to a solo dev like myself, and support any reasonably decent modern models (DeekSeek 4.x, GLM 5.x, etc). Fireworks has been a real dud, but Baseten in FAST mode has been absolutely amazing. Not cheap - but I find it substantially cheaper than Terra / Sol (low), etc. I haven't found Haiku to be smart enough to be useful.
DeepSeek (the provider) and beyond seem to hit the sweet spot to keep me mostly in-context, so I don't need 4 parallel work items to feel productive. So roughly 200tps or greater. Vanilla Baseten is ~400 tps, and FAST is 700-800 avg.
Price wise, Deepseek hosting is extremely cheap, Baseten is a step up - maybe 5x or more, and FAST mode is well beyond that. But I still find myself dual wielding Deepseek and Baseten FAST when I need the speed.
The other benefit: for my own work in very limited hours, I'm more than happy to pay for answers to come back faster.
Can you explain how please, in detail? I'm failing to see how something that needs a lot of hand holding (the AI we have), when sped up is somehow capable of taking our jobs.
Unless we get some kind of very different architecture I’m not sure we’re going to get a 10-100x improvement any time soon, api models will always preference throughput over individual task latency, so any improvement that sacrifices serial decode for better parallelism is going to be more economically viable.
While the LLM speed will increase, some stuff will be incompressible: running unit tests, tool use, UI interaction / mouse control. In my case, a significant amount of the is already spent on that, so much that I rarely ever select fast mode.
Pretty expensive nowadays, but it’s hard to think that prices won’t come down eventually. AI is being used everywhere, so I’d expect hardware investment to grow exponentially, including in the custom chips that power ultrafast models. To me it feels like we are in the "telephones are for the rich" era of AI history.
Yeah. This single accident was more deadly than all commercial air traffic in the EU in the past 10 years together, while 50 deaths per day is the average for car crashes in the EU. Wild orders of magnitude difference in safety.
Rough estimate: by car, you can travel from Lisbon to Talinn (4000km) 100,000 times before expecting to die; by train, almost 4 million times, by plane 400 million times.
And if you're the median person (not the average person), you can probably 3x those numbers, because such a large number of casualties in traffic are risk-seeking individuals (e.g. a drunk driver is included in "average person" but not in "median person")
Maybe, but the median person shares the road with those risk-seeking individuals. Drunk driving accidents often kill people other than the drunk driver.
Because it's a bullshit slop research. Under computer programmers there is "Develop Websites" at 0. Apparently AI cannot write web sites. Or just nobody looked at the slop they generated at Mercor.
The point GP was making is that DNS-record based ACME mechanism for verification of ownership implies trust in the domain name owner (or really, anyone who can edit records in the domain zone), making the issuance of a certificate actually superfluous.
DNSSEC is probably not even used, though I never checked.
Eg. as a domain owner, you will put a cryptographic hash into your DNS zone so a CA can validate you have control over it, and then issue a different cryptographic hash derived from their private key, and browser will use their public key to validate this cryptographic hash (TLS cert) is valid.
So we could simply push public key into a DNS zone, and browsers could use it to decrypt the traffic encrypted by the private key from the server hosted under that DNS name: no CAs needed, similar to SSH except the DNS-ownership-implied-trust component.
"Extended validation" certificates involved a lot more (in theory, checking true ownership, business address, physical presence, etc), but nobody really cares about these, and with the push to automated renewal and 45-day expiration dates by 2028, it's going to make even less sense.
Edit: I realize now that you may be referring to the fact that MITM DNS server can inject a different public key in there and thus DNSSEC is required — you are absolutely right, and this is a good and important point.
I wouldn't be surprised if it is in some places, but nevertheless I wonder - is there a good reason for that? I don't think any lawmaker actually wants tax calculations to have the potential of not halting, right? ... right?
I wonder what the experience is of people here with the ultrafast mode (8x) of Codex.
reply