storus

20 ms·

331.
by storus 9mo ago
Does Qwen3 allow adjusting context during an LLM call or does the housekeeping need to be done before/after each call but not when a single LLM call with multiple tool calls is in progress?
332.
by storus 9mo ago
I don't think so, those things happen when agent yields the control back at the end of its inference call, not during the active agent inference with multiple tool calls ongoing. These days an agent can finish the whole task with 1000s
333.
by storus 9mo ago
Why not replace the context tokens on the GPU during inference when they become no longer relevant? i.e. some tool reads a 50k token document, LLM processes it, so then just flush those document tokens out of active context, rebuild QKV cac
334.
by storus 9mo ago
This is pushed by Antropic, OpenAI doesn't seem to care much about "skills". Maybe Anthropic is doing some extra training to better follow sections of text marked as skill, who knows? Or you can just store what worked as a sk
335.
by storus 9mo ago
Doesn't this risk some unforeseen effects on Earth or the rest of the solar system at that scale? Disruption of magnetic shield, some not yet known law of physics suddenly getting felt etc.?
336.
by storus 9mo ago
What are the SOTA methods for context management assuming the agent runs with its tool calls without any break? Do you flush GPU tokens/adjust KV caches when you need to compress context by summarizing/logging some part?
337.
by storus 9mo ago
Basically yes. First figure out how to do reliable models that can be well-aligned with safety expectations, then invest into a startup that brings that tech to the industry.
338.
by storus 9mo ago
Wouldn't it be better just to stack functionalities of multiple agents into a single agent instead of getting this multi-agent overhead/failure? Many people in academia consider multi-agentic systems to be just an artifact of the
339.
by storus 9mo ago
Do I need to have two M3U 512GB MacStudios to run this?
340.
by storus 9mo ago
It seems like the worst practices from Trilogy/Crossover are leaking all over the industry. First the crunch at all times at FAANG, next tracking everyone in a few minute intervals, ending up with real-time video tracking at all times,
341.
by storus 9mo ago
We currently can't do "AI safety" even in bleeding edge alignment research so investing into some startups in that area is just burning money. Current LLMs/ViTs have non-zero probability of producing something unsafe and
342.
by storus 9mo ago
Anywhere in Germany? E.g. Frankfurt has much higher density of subway lines and trams.
343.
by storus 9mo ago
Again prioritizing low cardinality event (stalking) instead of high cardinality event (theft) because of "security", making the device mostly pointless, good only to quickly locate some thing at home (assuming battery still holds
344.
by storus 9mo ago
So they made it impossible to remove speaker, destroying its usability as a theft tracking device. One could add an airtag with a removed speaker on a bike/scooter/car and then localize it in case of theft. With the new airtag any
345.
by storus 9mo ago
When I switch between macOS users frequently, in 2-3 days I get pink garbage on the screen and/or the windows start literally falling apart and macOS at some point hangs with a rotating wheel. On a 128GB RAM M3 Max. This persisted acro
346.
by storus 9mo ago
Just today when trying to run Forza Horizon 5 from a new laptop I got an Xbox login popup with some QR code I was supposed to scan but the window was super small and I wasn't able to resize it to actually see the full QR code. One has
347.
by storus 9mo ago
Mint doesn't support latest hardware. I've just tried to install the very latest Mint released a few days ago on ASUS ProArt P16 with 5090 and it couldn't even boot the Mint installer, it just froze with the Mint logo on the
348.
by storus 9mo ago
OK, but how is this going to be taxed? That's where the problem is. Maybe it will be cheaper to incorporate but then what, are individual countries going to lose on their expected taxes?
349.
by storus 9mo ago
EU can easily pull the plug on Euro as a reserve currency if they confiscate the money of a certain country and give it to another one. That would be the "front fell off" moment for Euro.
350.
by storus 9mo ago
EU car manufacturers decided that cheaper-to-make electric cars must be sold as luxury vehicles and failed to achieve economies of scale China did with their underwhelming initial models they kept improving relentlessly every year.
351.
by storus 9mo ago
Stanford started doing 15 minute exams with ~12 questions to combat LLM use. OTOH I got a final project feedback from them that was clearly done by an LLM :shrug:
352.
by storus 9mo ago
Isn't Cho Oyu (around 8201m) the highest non-technical peak in the world? Ice wall can be often bypassed/traversed around. Aconcagua is ~1.25km shorter.
353.
by storus 9mo ago
> helping the interface stay out of the way, removing clutter so the actual content of the app takes focus instead. Yeah, like when I need to guess what is clickable and what isn't...
354.
by storus 9mo ago
My commercial agents are just wrappers on top of GPT/Claude/Gemini so the standard deployment ways on Azure/AWS/GCP apply with integrations to whatever systems customers have like JIRA, Confluence etc. Some need to autom
355.
by storus 9mo ago
I had something similar with GPT, like a clockwork every day after like 1pm it started producing total garbage. Not sure if our account was A/B tested or they just routed us to some brutal quantization of GPT, or even a completely diff
356.
by storus 9mo ago
Yeah, Exo was one of the first things to do - MacStudio has a decent throughput at the level of 3080, great for token generation and Sparks have decent compute, either for prefill or for running non-LLM models that need compute (segment any
357.
by storus 9mo ago
I have MacStudio with 512GB RAM, 2x DGX Spark and RTX 6000 Pro WS (planing to buy a few of those in Max-Q version next). I am wondering if we ever see local inference so "cheap" as we see it right now given RAM/SSD price tren
358.
by storus 9mo ago
M3 Ultra with DGX Spark is right now what M5 Ultra will be in who knows when. You can just buy those two, connect them together using Exo and have M5 Ultra performance/memory right away. Who knows what M5 Ultra will cost given RAM&#x2F
359.
by storus 9mo ago
Will they ban Tor as well? You can visit those sites with Tor easily anyway.
360.
by storus 9mo ago
Each new macOS version brings new restrictions causing some essential apps to stop working or work in a more complicated way so I keep delaying macOS upgrades as late as possible. macOS used to be an OS that lowered my cognitive complexity

More ›