nijave

9 ms·

181.
by nijave 4mo ago
>it takes it sweet time to get code rolling, not the fastest model by any means Which provider are you using? I got a z.ai Lite Coding Plan and it's my understanding z.ai is on the slower side of providers and the Lite plan gets low
182.
by nijave 4mo ago
Calling GLM-5.2 locally hostable is a bit of a stretch. It's 1.5Ti of weights at bf16. FP8 requires >800Gi of VRAM which is well into data center multi-GPU systems
183.
by nijave 4mo ago
Yes Monty started MySQL then forked when either Sun or Oracle acquired it. My, Maria, and Max are his kids. There is MaxScale which is a L7 SQL proxy/loadbalancer
184.
by nijave 4mo ago
1 & 2 you can do with k8s host network on pods and nodeports ECS is a simpler alternative if you're on AWS.
185.
by nijave 4mo ago
Absolutely. You can also run codeburn or ccusage and they'll scan the session files and tell you how much you burnt in API token pricing equivalent.
186.
by nijave 4mo ago
Well, we're probably not going to be running frontier models anytime soon, but I think the general assumption is smaller models will continue to improve until they're sufficiently good frontier models aren't needed. There&#x2
187.
by nijave 4mo ago
Also considering libcurl which is an excellent, feature rich, high performance, and battle tested http client.
188.
by nijave 4mo ago
Between RAM prices and tariffs (if you're in the US), cheap stuff got a lot less cheap in the last year Also curious what your use case is? They seem highly catered towards video applications versus more generic IoT like esp*
189.
by nijave 4mo ago
There is middle ground. Authn can be stateless and authz stateful. Usually it's impractical to shove all the authz nuance into a JWT anyway Doesn't address logging out a single session, though
190.
by nijave 4mo ago
Your database can double as the revocation list. You can use a last modified timestamp on the user or a monotonic counter to determine if a JWT is stale.
191.
by nijave 4mo ago
The word "session" is overloaded. In browser terminology, a session cookie is one that expires/is cleared when you exit the browser (it doesn't persist on disk) In application terminology, a session is user state that ou
192.
by nijave 4mo ago
Agree, snippet is conflating a lot of things. JWT is a signed JSON blob. Cookie is a storage and transport specification. Local storage is a storage spec. A "regular" cookie could also be a signed cookie which is basically the sam
193.
by nijave 4mo ago
The lockfile should protect you there. It'd only be an issue if you're working on updating dependencies in which case there's other protection like min-release-age If pulling down your company repo and running `npm install` c
194.
by nijave 4mo ago
Taking things down doesn't help much unless the platform has something in place to make it hard to recreate them. >they could do the hard job of combining leads and working with appropriate agencies to maybe find and prevent these t
195.
by nijave 4mo ago
>These tools with arbitrary code execution when trying to download some code have got to stop But you still end up with the code on your machine and risk it being ran. Bigger issue is giant, inscrutible dependency trees. In this example,
196.
by nijave 4mo ago
Haven't used either but have heard good things about PlanetScale and NeonDB. Neon has a free tier
197.
by nijave 4mo ago
That's the rub. There is no such thing as "anyone that can supply hardware cheaply" Either it's an established vendor with designs and fabs or it's a newcomer that needs to invest a massive pile of cash in designs a
198.
by nijave 4mo ago
AI says probably Rockchip RV1103B which is Luckfox Pico Mini
199.
by nijave 4mo ago
I'll respond with more anecdotal evidence, the Llama family has been terrible at following directions in all the tests I've done--not sure about the other models in RULER. In the Chroma results, they look at Sonnet 4 which was als
200.
by nijave 4mo ago
Shrooms, sometimes crack
201.
by nijave 4mo ago
>making basic recall mistakes at <100k tokens. I usually see this when the context gets "tainted" as I call it. The model gets stuck on a bad path and there's no way to bring it back without clearing the context and sta
202.
by nijave 4mo ago
Claude Code seems to automatically do this in some cases. It seems to have some heuristic "will eat a lot of context" where it decides to dispatch a sub agent. I see it pretty frequently in troubleshooting and data analysis flows
203.
by nijave 4mo ago
Maybe it is time to reintroduce agents to Extreme Programming. Mainly leaning into frequent integration (putting the continuous back in CI) Imo merge conflicts are usually a symptom of another problem like poor coordination, poor architectu
204.
by nijave 4mo ago
I suspect it's much worse with agents. They tend to generate a lot more code and also more likely to make large refactors. From the parent, making edits to different functions in the same file usually doesn't cause a merge conflic
205.
by nijave 4mo ago
I'm split on Django--it's ok but I don't love it. I don't think it's chaotic. I won't deny it's somewhat slow however usually anything performance sensitive gets shoved in a native code extension anyway. A
206.
by nijave 4mo ago
In Ohio (or at least my county) the deed and mortgage are public record. As is a record when the mortgage is paid off. Interestly also property tax charges and payments are, too
207.
by nijave 4mo ago
& many times the sources don't want to reveal their identity or go on record. A sort of tradeoff--to get the info they have to protect the source "You may not talk to the media" is pretty standard language in US employee
208.
by nijave 4mo ago
He might be able to rely on them, but can they rely on him? It's fully possible he consults them then completely misses or butchers the message (really I have no idea, I know very little about him)
209.
by nijave 4mo ago
Anthropic released a new class of model called Mythos a tier above the last one, Opus. The Mythos model was designed for cyber security then they tried to undo that (my understanding) for Fable So arguably "more dangerous" by desi
210.
by nijave 4mo ago
>The world is a bit bigger than US and China, if Anthropic did it, another company can do it as well. Well, yes but it also took questionable legality and a massive pile of cash to get there. OpenAI has raised $180b and Anthropic $132b D

More ›