Pure agentic loops with markdown documents as a program 'agentic workflow' is incredible for experimentation, developing and testing your workflow idea.
The second it works, bake the workflow into the harness.
Yesterday I was doing just that, and the whole agent loop disappeared because the process could've been condensed into a one-shot request (+1 MorphLLM fast apply) from careful context construction. (It was an Autoresearcher)
It’s that I try and then can’t. When stuck in bed I can feel this momentum building in my head to push for movement and the a surge of will and then nothing. I didn’t reach the threshold of exerting my will and now I’m waiting for the next wave.
> solar panel installed doesn’t disappear if China changes their stance
Most countries have days to, at most, months of imports of oil in reserve. In contrast, a panel embargo wouldn’t have disastrous effects for years. But reliance it is the same. If you’re dependent on Chinese panels, China can cap your energy growth at whim. The degradation will be slow thereafter, but present nevertheless.
Using foreign panels for anything other than bootstrapping domestic or allied production would be the EU repeating its follies first with Russia and then with American LNG.
Stopping new panels in some hypothetical scenario is very different than stopping fossil fuel delivery when ch can stop ongoing energy production - it not even in the same timescale of problem
What are the alternatives for Europe? Continue to import oil and gas? Have some of your most important economic inputs price and supply controlled by the dumbest egomaniacs alive?
Nuclear? Good luck building it on time and on budget. Also where exactly are you getting that Uranium from? I’m not necessarily against nuclear I just don’t think there’s much you can do in five or ten years to move the needle with Nuclear.
Wind? Actually a good option as it has a strong domestic supply chain.
Solar? Buy China’s cheap panels as long as they are selling. If they stop selling figure out how to do it yourself. It’s not some big mystery how panels get made, China just had the foresight to invest in the scale required to drive prices down.
Coal? I mean at least it’s local. But solar + batteries are either beating it now or will be in the next few years if the same trends that have held for the last 30 years continue for the next 2-5. So you’d be investing in a more expensive, dirtier technology for what end?
There is no world where you get to not make a decision and the risk just disappears. I think renewables have the clear advantage here and have very manageable risks.
> Also where exactly are you getting that Uranium from?
Uranium can be stockpiled relatively easily (france had 4-5 years of uranium stockpiled). Since it is about 1% of the energy cost, that’s pretty inexpensive.
Also, uranium comes from suppliers on 4 different continents, there is little chance that it becomes unavailable overnight.
You could do that, too, if you want. The relevant questions would be how much of the electricity cost is buying solar panels, how are they stored, and how they age when not used.
Stockpiling uranium addresses a specific risk related to sourcing in an inexpensive way, I'm not sure what problem you address by stockpiling unused solar panels, and for what cost.
Importantly, it used to be Germany which had all the expertise, until the CDU government destroyed much of the German solar industry over night. It's funny how everyone always talks about Germany stopping Nuclear energy but nobody ever talks about the fact that subsequent German governments destroyed the renewables industry twice (and they are talking about it again), largely due to lobbying from the coal, Nuclear and car industries. Definitely an interesting what if
Could you please send which lobbies worked on destroying renewables industry twice? (You probably mean destroying solar industry, wind industry is up and running).
I could only find that EU manufacturers of solar panels wanted tariffs on imported Chinese solar panels and EU builders and operators of solar power plants didn't want tariffs on imported Chinese solar panels.
There are solar panel manufacturers outside of China that have no dependence on Chinese inputs such as polysilicon, wafers, and ingots. Two that come to mind are First Solar (US) and Toyo Solar (Japan). I’m sure there are others. Europe can buy from them while scaling local manufacturing.
The EU has the talent to ramp local production of panels and batteries in years, which as the parent said is how long a panel or battery embargo would take to really cause a crisis.
I mean the EU has ASML, the Large Hadron Collider, and ITER, among other things. There is no engineering talent problem.
I’m more concerned that we do not have the supply chain. Like, sure, we have people who can build solar panels, but are the components local? I wouldn’t expect so, we would very likely import from china. Developing effective supply chains takes decades, it’s not really something you can do right away with the level of precision required by modern technology
Look at how fast various nations ramped up advanced (for the time) military production before and during WWII, or the Manhattan Project, or the Apollo program, or China's rapid rise.
Engineers who know how to build factories, batteries, and solar panels could sit down and create a "war plan" to build out and scale infrastructure quickly if you asked them to do it and then got out of the way.
The EU has plenty of talent with the know-how to do this. If it couldn't be done even in a crisis situation, that's a political problem.
it’s my understanding that inefficient bureaucracy is the biggest stumbling block for rapid infrastructure or technological growth. engineers can get it done but the bottle neck will likely be to do with how fast government bodies can move
panels themselves are highly simplified chip-like production. silicon crystals and some dopants. anyone can make extruded aluminum. anyone can build power electronics, make copper or aluminum wire.
the only interesting parts here from a supply chain perspective are power transistors. europeans have been known to design these, but idk how easy it would be to start producing them locally. they have macroscopic feature sizes though.
it would take several years of iteration to get a functioning pipeline that ran at volume, but none of this is hugely complicated. certainly not decades.
the real problem is financialization. you have to float that plant with the understanding that its not going to be competitive.
My Chinese built inverter functions just fine without Internet, but I'd have to take over doing what Amber Energy are doing if I lost access to the cloud.
But that's residential scale: at grid scale these things wouldn't be online in the same way anyway.
and what if they can use your inverter to destabilize not only your netwrok, but the local grit as well? and how far it can go if sychrnoized with malicious intend?
Not the panel itself, but the firmware of the solar panel charge controller and inverter that's connected to the Internet because there's an app to monitor the system. I wouldn't bet that there aren't remote kill switches deep inside that firmware.
"However, rogue communication devices not listed in product documents have been found in some Chinese solar power inverters by U.S experts who strip down equipment hooked up to grids to check for security issues, the two people said.
Over the past nine months, undocumented communication devices, including cellular radios, have also been found in some batteries from multiple Chinese suppliers, one of them said."
It would not suprise me if not only Chinese manufacurers did this. Cellular modems are cheap and and the capability to cause blackout is very usefull.
this is where the real risk is. nobody can stop you from directly tapping the panel’s power but an inverter can potentially be bricked if it has internet. this is more an issue with residential than industrial. i would hope that all industrial panels are air-gapped specifically to pre-empt this scenario
Power electronics shall never be connected to the "Internet".
Any such installations of solar panels, batteries and the like must be interconnected only in a private network without Internet access.
For remote monitoring and control a proxy mini-PC must be used, to which one should use an authenticated and encrypted connection.
For any competent person, this is trivial to do today, to ensure that even if some electronic device includes a backdoor for its vendor that backdoor cannot be accessed.
If there exists any kind of wireless connection provided by the vendor for a device, it must be disabled, e.g. by removing any internal or external antennas. Unlike wired connections that can be filtered externally, wireless connections cannot be secured.
There's no harm in a string, only in the execution.
I create Tools as Actors, which you preconfigured for the LLM context (in-house agent loop). The tools being preconfigured means you setup their environment before they can be executed. If it calls a bash tool for instance, the Tool Actor gets called and then it runs that command against an attached remote VM.
Or filesystem operations, are just read/writes inside a .zip file, which is overlayed onto the target project at build time.
This article is spot on, and I probably say that because it's self reinforcing.
Do you mostly just hit the session limits? If so I know it's not ideal but you could wait an hour or two for that to reset. Not sure if that would work for you but just a suggestion
Consumer CPUs don't have enough PCIE lanes to do that. Even if they had physical x16 slots, at most two of them would be x16.
What's cheap to you? You can find Epyc 7002/7003 boards on ebay in the $400 range and those will do it. That's probably the best deal for 4x PCIE 4.0 x16 and DDR4. Probably $500 range with a CPU. That's in the ballpark of a mid to high end consumer setup these days.
That's the path I'm taking.
All with all those PCI-e lanes available.
If I had more income, I would also buy 4x 96g Optane drives of p0 swap disks and a few ssd's for p1 swap disks. To evaluate how well you can get a 1T model running in these absurd ram prices.
Even if you're fine with risers, that might not be enough. If the bridge lanes are PCIe Gen 3, as many consumer boards have, your Gen 5 card might not init. I extensively tested several motherboards to try and get my AM5 CPU talking to a triple Radeon AI Pro 9700 XT setup, and they absolutely refuse to come up on PCIe3. I was using dummy EDID plugs for them, so they think they have a display, ruling out that issue.
What I eventually had to do was buy a used Threadripper box to run those cards, because PCIe Gen 4 definitely works.
I want to spend $1500 for a card that can run a proper large model, even if it only can do 25 tk/s.
Intel is squandering a golden opportunity to knee-cap AMD and Nvdia, under the totally delusional pretense that intel enterprise cards still have a fighting chance.
Can confirm doing so is awesome. Get some slightly bigger ones and partition them for additional use as zil. They're extremely satisfying to use, and depressing to remember that we'll never see their like again.
Sure! This is more or less how I'm using Optane in my storage box:
Two of U.2x4 to PCIe x16 riser cards, one loaded with 960GB Intel-branded Optanes, one with 1.5TB IBM-branded. PCIe bifurcation is set up in the BIOS to let them all come up properly, where they just show as regular NVMe. Riser cards like this can easily be substituted for PCIe to SAS/Oculink to U.2 cables, if that would be more accommodating to your chassis.
Once they all come up, partition them for your preferred split of swap and ZFS special. Swap should have them all mounted with the same priority and discard=pages, I also recommend setting up zswap (not zram swap) with lz4 as an additional layer of fast, evictable memory pool, as well as `vm.overcommit_memory=2` and `vm.swappiness=150`. This will effectively give you really good memory tiering for workloads and file cache.
When adding the other partitions to ZFS, use `-ashift=12 special mirror dev dev special mirror dev dev ...`. ZFS special covers all metadata, the intent log (sorta write cache), and optionally small files. I like to set it up so <= 8k small files get sent there, but you can probably go higher depending on how much capacity you allocate. My ~24T of allocated data ended up being ~150GB special with 8k small file, and that's with the whole pool configured with deduplication and blake3 for all hashes. Blake3 is fast as heck, but has very long hashes, so from a metadata standpoint, I'm using the most expensive option. I mitigate that a bit my setting metadata redundancy to `some`, since my metadata is effectively RAID10 anyway.
With some extra NVMe/Optane allocated to regular ZFS read cache, and all my spinning-rust data VDEVs also as RAID10, it's almost like having the whole array in memory, or at least on fast flash. Eliminating metadata from your drives seeking and letting them be written nearly instantly with Optane does wonderful things for spinning rust :)
The second it works, bake the workflow into the harness. Yesterday I was doing just that, and the whole agent loop disappeared because the process could've been condensed into a one-shot request (+1 MorphLLM fast apply) from careful context construction. (It was an Autoresearcher)
reply