Slava Solodkiy.cv (@SlavaSolodkiy) header
Slava Solodkiy.cv (@SlavaSolodkiy) profile picture

Head of Curiosity @ digital https://xitter.cf/t.co/upokfRUZ8p x https://xitter.cf/t.co/T2Q5nOwyww l Founder and ex-CEO @ https://xitter.cf/t.co/3ZfFRnfUyv x https://xitter.cf/t.co/mAVCshMSIH l https://xitter.cf/t.co/hdLDJUAXvC

Singapore - London - LA
Joined August 2020
Photo by Slava Solodkiy.cv (@SlavaSolodkiy)Photo by Slava Solodkiy.cv (@SlavaSolodkiy)Photo by Slava Solodkiy.cv (@SlavaSolodkiy)Photo by Slava Solodkiy.cv (@SlavaSolodkiy)Photo by Slava Solodkiy.cv (@SlavaSolodkiy)Photo by Slava Solodkiy.cv (@SlavaSolodkiy)Photo by Slava Solodkiy.cv (@SlavaSolodkiy)Photo by Slava Solodkiy.cv (@SlavaSolodkiy)Photo by Slava Solodkiy.cv (@SlavaSolodkiy)Photo by Slava Solodkiy.cv (@SlavaSolodkiy)
The Atom's Fizzle: How We Learned to Stop Worrying and Love… Solar Panels? medium.com/@slavasolodkiy_67… Ah, the Atomic Age! That glorious epoch that never quite was. Picture it: a utopia powered by miniature suns, cars zipping around on vitamin-sized pellets of pure energy, and maybe, just maybe, a three-eyed fish or two as a fun little bonus. It was the future we were promised, a future so bright, we'd all need to wear those snazzy atomic-themed sunglasses (even indoors). Instead, we got... well, reality. A reality where the only thing glowing brighter than a reactor core is the smug smile of a solar panel salesman. So, what's the moral of this atomic-sized cautionary tale? Perhaps it's that sometimes, the shiny, futuristic dream isn't all it's cracked up to be. Sometimes, the simpler, less explosive solution is the better one. Or maybe it's just that we should all invest in solar panels and start practicing our smug smiles. After all, in the energy game, as in life, it's always nice to be on the winning side, especially when the losing side has a tendency to go boom. And by the way, don't forget to install an underground bunker in your backyard just in case. You know... for old times' sake. #NuclearEnergy
2
1
5
1,117
We're giving away a Hogwarts Legacy custom wrapped NVIDIA GeForce RTX 5090 to celebrate its official release with DLSS 5 🪄 To enter: 🟢 Share this post 🟢 Comment #RTXON
31
Two @JollaHQ #Linux-based phones arrived 😍 works well w Spark via terminal 🚃 back and forth @SailfishOSNews
#DGXSparkFE @NVIDIAAI arrived on 22 Oct 🔥 and tested 🎇 works well btw w @OlaresOS too 🚀
38
Two @TiinyAILab arrived on 4 Nov 🚀🚀 finallyyyyy 💕
#DGXSparkFE @NVIDIAAI arrived on 22 Oct 🔥 and tested 🎇 works well btw w @OlaresOS too 🚀
149
#DGXSparkFE @NVIDIAAI arrived on 22 Oct 🔥 and tested 🎇 works well btw w @OlaresOS too 🚀
2
329
tested: @Intel @Arc B580 + @OlaresOS (w some bugs: github.com/beclab/Olares/iss…) + OCuLink: Not just GPU detection. Kubernetes recognised the GPU, OpenVINO served a reranker via HTTP, with ~33 ms average latency in my test. A modest but real end-to-end result for #localAI. 12GB->36GB?
a fairly mundane question: (what would I actually do if two more boxes arrived — and) if I could borrow two more kinds of silicon solodkiy.cv/aorus-intel-tiin… (I do not own yet, only manifestating)? →can compute move between machines, vendors and power bills, while the right to decide stays put? □ Two @TiinyAILab Pocket Labs are already in transit. Not so I can screenshot 120B in a tin. Because they are two separate physical objects. I can delay one, unplug one, replay an old packet, skew a clock, and see whether a mandate still holds. ■ The rest of the bench is already here: @Olares_OS One — mobile @nvidia 5090, 24GB, Thunderbolt 5. The CUDA reference that already ran Nemotron-FOL and the Sovereign Decision Plane. @Hi_MINISFORUM M1 Lite — 96GB, AMD 24GB. @iMoreFine M900 — 48GB, NVIDIA 12GB. The box FreeToken first used to put a 35B MoE where the brochure said it should not fit. □ #Wishlist → @AorusOfficial RTX 5090 AI Box (good price w @ScanComputers btw) — desktop 5090, 32GB, TB5, sitting next to the 5090 already inside the Olares. The only honest question is whether eight extra gigabytes unlock a named class of work that 24GB cannot do cleanly — second reads, long context, less brutal quant, multimodal state that has to stay resident. If the list is empty, that is the result. (The #PerfectInfinityCraft card over OCuLink is the same question on a different bus: dram.gold/aorus.html redesign note, not a review.) @intel Arc Pro — @MaxsunOfficial Dual B60 first only if the host actually exposes both x8 links (it is two 24GB chips, not a magic 48GB heap). Cleaner first Intel target: a B70 32GB (GUNNIR / MAXSUN class) on the Morefine or a verified workstation. Not “Xe2 versus CUDA”. The question is whether the decision pipeline is portable, or secretly a CUDA application with better copy. ■ The six experiments I would run, in order, and publish including the ones that fail: [1] Authority under partition. Perception on one machine, verification on another, deferred evidence late from a third. Caps, TTL, revocation, ai_authorised_to_override = false. If a delayed FAIL can widen a mandate, the architecture is theatre. [2] Are workers fungible? Same evidence packet through CUDA, Tiiny runtime, later XPU. I do not need identical prose. I need the same permitted action — or an escalation when the backends disagree. [3] Smallest sufficient perception. Start cheap. Escalate only when a gate says the first read was not enough. Metric: fraction of real tasks finished correctly at each rung, not tok/s. [4] AI Box as capability, not speed. Named workloads that are awkward in 24GB and clean in 32GB over TB5. Plus the Olares eGPU caveats: Gen1 workaround on Olares OS, Gen4 after restart on Windows. The bus is part of the measurement. [5] Software sovereignty. Port the real stack — packets, Prolog/Z3/OPA, regression cases — onto Intel. Installation pain is data if “sovereign” is the claim. [6] Power → intelligence. Only after 2 says which jobs are safe to move. Then schedule by joules and pence per verified decision. A cheap wrong answer is not efficient. I am not trying to build a leaderboard in a flat in London:) ■ I want a public harness other developers can rerun: stale state, delayed FAIL, replayed packet, clock skew, node loss, TTL expiry. Raw configs. → Verdicts separate from the model’s explanation. A negative-results folder that does not get deleted because it made a favourite box look unnecessary. This week started as a hardware wishlist - it is ending as the same Decision Plane, asked across more than one machine: → let AI perceive and explain; put truth, policy and authority somewhere else; then check whether that “somewhere else” still works when the workers are cheap, late, or a different vendor. Not “world first”. A working note. 'What if?' questions first. Hardware second. Failures included. @AORUS_UK @GigabyteUK @nvidia @NVIDIAAI @nvidiadeveloper continues...
1
1
111
2 tiny @SipeedLab NanoKVMs (Go and Go+). 2 physical @OlaresOS-based PCs (@Hi_MINISFORUM & @Acemagic_MiniPC). 1 Tailscale network. I accessed both Linux consoles remotely from outside: independently of the host OS. All good - Mouse HID: nope. Next: AI-assisted bare-metal recovery.
1
1
533
One less technical memory from the evening: Jensen was asking people near the stage whether @nvidia had invested in them. When he got to me, I said “Oops, no!” His reply, as I remember: “Then I will!” I’m filing that under excellent stage improvisation, not a funding announcement
#JensenHuang speaks in Berlin tomorrow :: slavasolodkiy.substack.com/p… My ears are still in London:( I was supposed to be there. I won't make it — but hearing him twice on 14 October left me with four questions worth taking into the next keynote :: linkedin.com/pulse/listening… These are my recollections and interpretations, not a transcript. @nvidia people and fellow attendees: please correct or add to them!:) 1. YOUR AGENT NEEDS MORE THAN A PASSPORT. Who created it? Who is it acting for? Who can change its role or revoke its mandate? What evidence travels with it into somebody else's system? Identity is not authority. Knowing who someone is does not authorise them to move your money. That's the question behind my AuthorityClaw experiment: solodkiy.cv/authorityclaw.ht… 2. THE NEXT AI EXPERT MAY BE AN EXPERT IN SOMETHING ELSE. I heard a shift from tools for creators and developers towards agents adapted to particular industries. Less “How do I become an AI specialist?” More “What important problem in my industry remains unresolved?” The customer doesn't wake up wanting an agent. They wake up wanting something to work. 3. YOUR NEXT PC MAY NOT BE WITH YOU. OR ENTIRELY YOURS. First you went to the computer. Then you carried it. Next, perhaps you simply reach it — from whatever device is handy. That resonates with my @Olares_OS experiments. My next question is sharing: perhaps ownership matters less than authorised access, privacy and control. solodkiy.cv/olares-seamless.… 4. THE AI FACTORY NEEDS NEIGHBOURS, NOT JUST MEGAWATTS. At the evening event, I remember openness to different energy sources and infrastructure integrated with nearby communities. That connects to a hypothesis I'm developing: a Power Node combining electricity, compute, grid flexibility and useful local services. solodkiy.cv/power-to-intelli… My Hoxton Jobcentre concept asks what the last part could mean for people looking for work: solodkiy.cv/hoxton-jobcenter… What happens outside the fence matters too. Four subjects. One recurring question: WHO GETS TO USE THIS CAPABILITY, ON WHOSE AUTHORITY, AND FOR WHOSE BENEFIT? Thanks to Elliot, Ayesha, Mozhgan and Manel for London and the Berlin invitation! Sorry to miss the latter:( My curiosity hasn't cancelled its trip. @NVIDIAAI's own London recap: blogs.nvidia.com/gtc-berlin-… Those listening tomorrow: what carries through, what changes, and what did I miss? My session shortlist is in the comment below. What question are YOU taking into the room? Keep asking linkedin.com/pulse/listening… slavasolodkiy.substack.com/p…
1
1
96
waiting for my @TiinyAILab prosumer.one/guides/tiiny-po… delivery - and counting days!) as a kid on Christmas Eve, hahah
Buy TWO #AI gadgets on @Kickstarter: Wake up with a secret THIRD one for free:) @TiinyAILab Tiiny Pocket Lab does the heavy uncensored thinking. @DecoKeeTech Decokee Quake becomes its face, hands and soul. No laptop. No cloud. Just a book-sized offline AI station that actually belongs to you. I already explained the cheat code 👇 open.substack.com/pub/slavas…
1
1
170
I’m a product guy who can speak fluently with engineers, translate messy business solodkiy.cv/tech.html and regulatory problems into technical specifications, and use AI-native tools to get working systems built and tested.
If you're building something where this matters — or you know someone who is — I'd appreciate a share lnkd.in/p/eq6A6J8K or an intro via slava@dram.gold I built a bank from scratch. Not as a metaphor — an actual US banking licence (OCIF IFE-065, 2021), $25M raised, 60 people, 26 countries, $150M AUM. Zero marketing budget the entire time. The last two years I spent thinking — researching digital identity, AI infrastructure, and the future of money. Writing books is how I prepare for things that matter to me. Now I'm ready to build again. I'm looking for a mandate: a bet to own end-to-end, inside a company that's serious about where fintech, digital identity, or AI is going. Or a fractional/advisory engagement if the problem is genuinely hard. London/Edinburgh/remote. Available immediately. Global Talent Visa → ILR → no sponsorship needed.
2
32
If the public, journalists and commentators are going to spend their attention on this incident, the most productive question is not “who screwed up?” but “what assumptions failed, and how could we rebuild them?” solodkiy.cv/revolut-incident… The lesson is ours, not just @Revolut's.
The domain was real. That is the part people keep skipping, because it ruins the plot :: revolut.hot/revolut-incident… If the story is “a fintech got hacked,” the crowd already knows how to sit. Popcorn, screenshot, moral. If the story is “a technically authentic government-looking channel carried an unauthorised instruction, and a disclosure desk did what disclosure desks are built to do — move,” then you are no longer watching a villain. You are watching a ritual that almost every regulated company still treats as plumbing :: linkedin.com/pulse/domain-re… @Revolut said the important quiet thing first: systems and customer funds were not the scene. The scene was a request that arrived through a legitimate agency domain, passed internal checks, and produced an irreversible event. Passports, licences, contact data; reporting also points at selfies, statements, transaction history. Once that packet leaves the building, “we blocked the sender” is hygiene. It is not un-send. The first public spine of that story was not a leak dump. It was reporting. @JagmeetS13 at TechCrunch put Revolut on the record and read the customer notice. @Reuters moved the same confirmation onto the wire. @mapamijares at Bloomberg carried it into the finance stack. @zachxbt did what desks often cannot: he put the notice in front of people who actually live inside these apps, and argued the targeting looked concentrated, not industrial. @MagicalTux said the quieter, uglier sentence from inside the affected set: they should not have sent the file. Secondary desks then did the necessary amplification — @DishitaInCrypto, @hardeyjumoh, @securityaffairs — each adding a shard: Bitcoin histories, the notice taxonomy, the fact that domain authentication is not a person. This is not a roll call so they can be dunked. It is an invitation. They already did the hard part: they established that the channel looked like the state and the authority was not. The useful next hour is not “which brand is finally exposed.” It is: what should have had to be true before a file was allowed to leave. I do not enjoy this as sport. I have spent too many years trying to make banks boring in the places where boredom is a virtue. Revolut is not a cautionary sticker to me. It is a company that made speed and surface area into a product, and therefore became interesting enough to attack with a method that is already industrialised: emergency-looking legal process, riding a channel that looks like the state. That method is older than this week. Compromised police mailboxes. Forged urgency. Kits. A market for “make it look official.” The new part is not the trick. The new part is how clearly it shows the missing object in most compliance stacks. Banks spent a decade teaching machines to recognise customers. They spent much less time teaching machines — and desks — what a requester of the state actually is. Customer onboarding has states, documents, fallback, step-up, “come back tomorrow.” Legal disclosure often still has an inbox, a tone of emergency, and a cultural terror of being the person who delayed a real warrant. I know that terror from the other side of the glass. At Arival I tried to stop treating the regulator as weather. Weather arrives, you issue a PDF, you hope. I wanted the agency to be a product user: an organisation account, named humans inside it, questions hanging off a customer or a payment rather than off a shared mailbox, a clock on the request, even the unromantic telemetry of who opened what and for how long. The bet was not “give them a search box and go home.” The bet was: if oversight is going to exist, stop making it an adversarial scavenger hunt. Some of that aged well. Interaction is a product. Actors, objects, latency, evidence. A named officer is better than disclosures@. One assumption did not survive this week, and I should say that before anyone copies the old sketch. We still treated a working government email as a reasonable root of identity. Convenient. Empirically insufficient. A live domain can prove a path. It does not prove that the human on the other end currently holds capacity, that the mandate covers this subject, that the scope is this field list and not “the file.” If you hand a compromised-but-real mailbox a read-only window into a bank, you have not modernised supervision. You have built a nicer exfiltration UI. That is why my first “smart” fix after reading the incident was also wrong. I caught myself writing the civilized version of the same leak: give a minimum dataset now, verify the person later. It sounds like adult risk management. It is the same class of failure with a smaller envelope. An address is not provisional once it has left. A phone number is not a rehearsal. “Less PII” is still PII. The useful embarrassment took about forty-eight hours. Kill the rule. Do not edit it quietly. What I would protect now is a harder line. A request may open a case immediately. A case may freeze deletion, assemble a packet that never leaves the building, escalate, start a clock. That is HOLD. HOLD is not disclosure. Disclosure starts only when five different questions have answers that are not “the email looked right”: the channel is authentic; the principal is a named person verified off that channel; their office is current; the mandate exists as an object — basis, case, jurisdiction, purpose — not as atmosphere; the scope is a list of fields, default empty. Until those are green, the outside world gets nothing it can use to impersonate a customer at another bank on Tuesday. None of this requires a religion about AI. A model can read a warrant and propose a scope. It must not be allowed to skip a red gate because the prose sounded like a ticking bomb. Urgency is a claim. Claims are evidence. Evidence is not authority. If I were in a Revolut building on Monday — and I am not applying for the job in this sentence — I would not start with a white paper. I would start with five ugly operational questions for the disclosure desk: Can a government-looking email, alone, move customer data? Is the human checked on a number or registry you already possessed, not one supplied in the thread? Do you check that their role is still live? Is the legal basis a structured object, or a PDF vibe? Is the default export the whole file, because “they asked”? If the honest answers are the unsafe ones, you do not have a communications problem. You have a state machine that jumps from “received” to “sent” without a revocable middle. The public conversation keeps asking the wrong upgrade. Better filters. Better staff. Better apologies. Fine. Also insufficient. The upgrade is to stop letting a convincing artefact spend operational power it has not earned. So if @JagmeetS13, @zachxbt, @mapamijares, @MagicalTux and the wire desks already spent the weekend establishing what left, the more interesting follow is what should have been required before it could leave. Not as a pile-on. As a design question that outlives this brand and this inbox. Revolut can treat this week as a brand wound. Or as a chance to publish the missing object: a disclosure warrant that is scoped, timed, dual-controlled on the irreversible fields, hashed with the rule version that authorised it, and unable to form if the principal is still a mailbox. I already built a softer version of the surrounding product, and I would not ship that version again. That is the only authority I am claiming. Not that I would have stopped their incident. That I know which of my own ideas would have made a similar incident worse, and which line I am no longer willing to blur. The domain can be real. The officer still has to exist. The file does not get to leave while we are being polite.
1
50
I built an evidence-gated app with @NVIDIA's #Nemotron :: lnkd.in/p/e3U9ippg 🟢 GREEN control: ALLOW. 🔴 RED model output: ESCALATE_TO_HUMAN. 🟦 #AI override: false. Synthetic data; captured-evidence replay. thnx @johnnync13 #NVIDIAGTC
2
68
saved all my experiments & conclusions with my next steps here: solodkiy.cv/FreeToken.html thnx a lot for David @TeksEdge and his kind advice to try #FreeToken looks very useful for #sovereignAI #localAI - and my @Olares_OS One, @Hi_MINISFORUM M1 Lite & @iMoreFine M900 w @AMD stack
#FreeToken experiment step#4: how low can we go? After @TeksEdge great post - testing FreeToken on: #1(smaller)(bigger model) on Olares One — 96GB RAM + 24GB NVIDIA VRAM #3 Minisforum M1 Lite — 96GB RAM + @ASRockInfo 24GB AMD eGPU via OCuLink (with mixed success on the unofficial ROCm path) …I wanted to try something much less comfortable: #4 @iMoreFine M900 • 48GB RAM • @ASUS @NVIDIAGeForce RTX 3060 12GB eGPU via @Hi_MINISFORUM DEG1 OCuLink • Windows 11 First surprise: FreeToken installed cleanly and gpt-oss-20b ran at 13.07 tok/s, using 9.51GB VRAM. But that’s not really the interesting test. So we loaded the official Qwen3.6-35B-A3B-NVFP4 — a 35B MoE model with only ~3B parameters active per token — and deliberately limited FreeToken to --memory-ratio 0.60. In other words, although the @ASUS_ROG RTX 3060 has 12GB VRAM, FreeToken was allowed to use only about 7GB of it — roughly recreating the kind of memory constraint from the FreeToken authors’ 8GB RTX 4060 Laptop experiment. Result: it works. • Qwen3.6-35B-A3B-NVFP4 • 20.53 tok/s decode • 189.5 tok/s prefill • ~7.07GB total VRAM used • 255-token completion successfully generated • Hybrid MoE backend selected automatically And this is where FreeToken becomes genuinely interesting to me. The 35B model did not somehow “fit” into 7GB VRAM. The expert bank lives in system RAM. FreeToken keeps a cache of useful experts on the GPU and decides what to fetch over PCIe vs what to execute on CPU. On this machine it measured: • PCIe expert gather: 3.3 GB/s • CPU expert bandwidth: 19.9 GB/s …and automatically chose a hybrid split of roughly: 14.8% PCIe fetch / 85.2% CPU execution. That also explains why we’re slower than the FreeToken paper’s ~39.3 tok/s on an RTX 4060 Laptop 8GB: their machine had dramatically faster host-side bandwidth = The bottleneck here isn’t really the 12GB GPU — it’s the path between RAM, CPU and GPU. Which, to me, is the important result though. We tend to classify local AI machines by VRAM: 8GB, 12GB, 16GB, 24GB… but FreeToken makes that taxonomy much less useful for MoE models. A fairly ordinary old RTX 3060, using only ~7GB of VRAM, just generated from a 35B MoE model locally at ~20 tokens/sec = (No magic “35B fits into 7GB”, sure) just a different memory architecture. And that may make a lot of supposedly “too small for serious local AI” PCs much more interesting than they looked a month ago. Next question: how cheap can the machine get before this stops being useful?)
1
2
143
I’m opening this mandate/job/EiR-role search lnkd.in/p/ec_35Zfa if there’s a problem on your side where I could be useful, I’d genuinely like to talk slava@dram.gold → If this resonates with anyone in your orbit, feel free to forward/share
1
1
1
41
If you have read #FlorianIllies’s '1913: The Year Before the Storm', you already know (feel!) how: Some things cannot simply be told. They can be played on a violin, can be painted, can be fucked. They can be written the way 'The Dark Side of the Earth' is written by @zygaro
как часто\обычно бывает - я начал не с того конца: будучи давно, и вполне себе даже наивно-искренне-восторженно, подробным читателем и поклонником каждой из книг @zygaro - я, узнав первый самый ещё раз о выходе "Тёмной стороны Земли", само собой, перво-наперво намеревался её прочитать как опубликуют - но совершил банальную ошибку: я прочитал рецензию на неё (кстати, вполне возможно, что она была от @meduzaproject , но это не так важно - все были хвалебные), потом вторую, потом третью. и каждый раз я всё дальше откладывал прочтение. (спойлер: именно этот твит Медузы в итоге-таки "заставил" меня взять-сесть-открыть-прочитать - ну и, как это всегда с книгами Зыгаря у меня: начал читать - и потом только разве что за чаем, да в уборную, - не могу остановиться, пока не закончится) а проблема (для меня - так что, вполне возможно, это вообще моя личная проблема) рецензий всех хвалебных этих была в том - что все они были "про ЧТО", а книга-то вообще не про это, она "про КАК" (на первых страницах упоминается Мамардашвили - так вот, у него много про это: не так важно что конкретно помыслили - важнее, как вообще мыслили, кто-когда и в каких условиях\контексте, как эта мысль цвела и текла, как это было красиво в моменте, в процессе). ну если вы читали вдруг "1913. Лето целого века" Флориана Иллиеса - и я бы вам вдруг стал рассказывать (хвалебно), что книга мол про исторические события пред-военного времени хронологически-дотошно рассмотренные под микроскопом автором, который будто бы взял у каждого из участников событий персональное (из 100+) интервью перед публикацией. ну Иллиес бы меня точно нахуй послал за такую хвалебную рекомендацию - так и у меня с рецензиями этими вышел конфуз. т.е. никто в этих рецензиях, конечно, не соврал, и уж точно ничего плохого не имел ввиду - просто книга Зыгаря вообще не об этом. (и, кстати, то, что она продолжается в форме театральных постановок - куда ближе к описанию!) тут сам формат, форма, ритм, пойманные недосказанности, сделанные нарочито или невзначай намёки и отсылки с пасхалочками (как от самого автора, так и от истории) - это ж вот всё гораздо важнее. нельзя же некоторые вещи просто взять и рассказать - их можно сыграть на скрипке, их можно потрахаться, их можно испечь, их можно нарисовать... их можно написать было как "Тёмную сторону Земли", и, наверное, их можно было только так
52
One small consolation: I nearly backed #Orcanyx, another remarkably similar tiny eGPU on @Kickstarter. After already backing multiple eGryphons, I decided not to. That decision is aging rather well & now I’m genuinely curious: are #eGryphon and Orcanyx completely unrelated teams?
I backed the #eGryphon portable RTX eGPU on @Kickstarter (promoted by #KickstarterPledges aka #BackerPledge and #PledgeBox) — initially 2 units, then several more specifically to help the project close its funding shortfall. But the Aug 18 backers update #21 kickstarter.com/projects/egr… is concerning enough that I think it deserves some daylight. eGryphon says its manufacturer has now refused to proceed with production, arguing that an earlier missed payment deadline invalidated their agreement. According to eGryphon, the manufacturer may return the most recent payment while attempting to retain earlier deposits/payments. eGryphon is now considering legal action, emergency financing, moving production elsewhere — or winding the project down and refunding whatever funds remain. None of this proves #fraud. But it does mean the project has moved well beyond an ordinary Kickstarter delay. The central question now is very simple: Where are the money and the hardware? Backers need a redacted but auditable account of: — payments made to the manufacturer; — money currently recoverable; — ownership/location of GPUs and other components; — existing inventory; — the relevant contract/payment deadlines; — remaining eGryphon cash; — capital required to restart production elsewhere. One line in the update is particularly troubling: if the project is wound down, eGryphon says refunds would give priority to “supportive positive backers.” I could be considered as "positive" or not - but that needs immediate clarification. Refund treatment should obviously not depend on whether a backer publicly supported or criticised the creator. I still hope the manufacturer dispute is exactly what eGryphon says it is and that the product gets made. But at this stage transparency is more valuable than another promise or revised shipping date. Show the paper trail.
1
6
1,112
Maestro @Chris_Skinner 's new bestseller 🔥 (and completely new style/format/genre!) 🚀 A razor-sharp, timely financial thriller that feels all too real amzn.eu/084XJaG3
2
3
5
342
Still waiting and waiting… looks like I’ll never get + the #eGryphon eGPU, #MeetFRIDAY NAS by #Ainest, and #KeyGo Gen2 Pro, - from @Kickstarter :( #KickstarterPledges aka #BackerPledge #PledgeBox
1
3
134
Nice one, @TiinyAILab × #Odysseus = demo is one of the more practical things I’ve seen lately in local AI! I’ve been running a mix of @Olares_OS + @gmktecdotcom + @softwincn's GPD Pocket 4 + eGPUs and various mini-PCs for a while. Ordered two Tiiny units specifically for this!
Tiiny x Odysseus: From Ideas to Action Shared by our community, we had to give Odysseus a try 🙋 With Tiiny AI Pocket powering the models, Odysseus helps turn ideas into structured documents and actionable workflows. See it in action 👇 #TiinyAI #AI #AIWorkflow #Odysseus
1
1
3
660
Having known @Chris_Skinner & his voice in #fintech for years: fascinating to see him channel that expertise into such compelling fiction amzn.to/4vz3Sde as 'Diary of a Ponzi Scheme' (Token Money, Token Love, Token Lies), a razor-sharp thriller that feels all too real
2
1
3
1,559