Homebrew AI Super-science. If you sacrifice honor for peace, you will have neither. #NAFO ** Π‘Π»Π°Π²Π°Π£ΠΊΡ€Π°Ρ—Π½Ρ–

Gainesville, FL
Joined February 2022
Now that Mike Lee has established that NAFO is a branch of the CIA, this for Senate hearings.
Someone’s behind the pro-Ukraine troll campaign known as β€œNAFO.” Who is it? A few possible candidates have been suggested. Which one seems most likely to you? thegrayzone.com/2022/10/28/s…
10%Pentagon Contractor
44%CIA Contractor
21%Ukraine (using U.S. $)
25%Soros
34,337 votes β€’ Final results
11
39
4
241
20,893
Allen πŸ‡ΊπŸ‡¦πŸ‡ΊπŸ‡ΈπŸ‡¨πŸ‡¦πŸ‡ͺπŸ‡ΊπŸ‡¨πŸ‡­ Fella retweeted
As Trump continues to bend to Putin, he is putting Ukraine at greater risk in its fight for freedom by restricting access to the technology Ukraine uses to intercept Russian missile attacks, putting untold Ukrainian lives at risk and undermining Ukraine’s ability to defend itself. We are demanding answers from SpaceX about who ordered these restrictions and why.
331
598
35
2,412
41,538
Allen πŸ‡ΊπŸ‡¦πŸ‡ΊπŸ‡ΈπŸ‡¨πŸ‡¦πŸ‡ͺπŸ‡ΊπŸ‡¨πŸ‡­ Fella retweeted
I will be speaking with Ukrainian President Volodymyr Zelenskyy and his team this week to discuss Starshield, sanctions, and other matters pertaining to the defense of our strong ally, Ukraine. I will be assuring President Zelenskyy that anyone who attempts to limit the use of Starshield for Ukraine will be fully investigated and will be held fully accountable, to include public congressional oversight hearings, and immediate harsh and punitive budget cuts to any responsible agency or agency head. Congress will unequivocally reaffirm that Ukraine is an indispensable ally to the United States, that Russia is a self-declared enemy of the United States, and that the United States will 100% support and defend the people of Ukraine in their fight for their own freedom against this brutal, illegal, and unprovoked Russian invasion. The United States must provide any and all defensive equipment that Ukraine needs to defend itself. The United States must not import a single drop of Russian oil. Sanctions against the Russian regime must be harsher and stronger than they’ve ever been before. Each and every one of these items will be brought to the House Floorβ€”and each and every one of these items will pass on the House Floor by an overwhelming margin. This is, after all, what Ronald Reagan would do. And this is what we will do now. Ukraine can, and will, defeat Russia. #PeaceThroughStrength
7,327
4,299
614
23,316
1,195,994
Allen πŸ‡ΊπŸ‡¦πŸ‡ΊπŸ‡ΈπŸ‡¨πŸ‡¦πŸ‡ͺπŸ‡ΊπŸ‡¨πŸ‡­ Fella retweeted
BREAKING: The diesel fuel Russia is now exporting to the United States at U.S. President Trump’s request may have originated in Iran, as Russian oil refineries are largely unable to produce diesel because of Ukrainian attacks. Russia has been importing most of the diesel it needs from Iran. In September, thousands of Iranian tanker trucks were spotted in various regions of Russia, including Ryazan, carrying Iranian diesel. Russia may be buying Iranian diesel and reselling it to the United States. If so, Trump’s attempt to lower U.S. energy prices and deceive American voters is putting money into the hands of Russia and Iran’s Islamic regime.
410
4,364
578
12,002
823,974
Allen πŸ‡ΊπŸ‡¦πŸ‡ΊπŸ‡ΈπŸ‡¨πŸ‡¦πŸ‡ͺπŸ‡ΊπŸ‡¨πŸ‡­ Fella retweeted
Ukraine's new campaign of destroying Russia's data centers will prove devastating for Putin's economy and war machine if the White House is unable to prevent it. The chaos we're already seeing across Russia is only four days into the campaign and 3 facilities targeted.
226
1,635
97
7,744
376,575
I have only used this for about 15 minutes so far, but it is the best terminal app I have ever come accross: waveterm.dev/ @wavetermdev
1
1
194
wave has both tabs, panes, and multiple windows. It can open and edit files from the target on the host with a simple command:
1
1
92
Allen πŸ‡ΊπŸ‡¦πŸ‡ΊπŸ‡ΈπŸ‡¨πŸ‡¦πŸ‡ͺπŸ‡ΊπŸ‡¨πŸ‡­ Fella retweeted
PrismML’s Bonsai 2 claims it matches 27B while 9x smaller, which would completely change how we value older GPU. Claim this big, I had to measure for myself, and it’s impressive with some shortcomings 27B level reasoning & math, 12B level coding & agentics in 7GB body.
22
12
2
175
13,415
WTF is this MVID company??? Why is it green??? Time for a personal visit from @414magyarbirds Headquarters coordinates: 55.699000, 37.343029 en.wikipedia.org/wiki/M.vide…
You’re unable to view this Post because this account owner limits who can view their Posts. Learn more
1
179
Allen πŸ‡ΊπŸ‡¦πŸ‡ΊπŸ‡ΈπŸ‡¨πŸ‡¦πŸ‡ͺπŸ‡ΊπŸ‡¨πŸ‡­ Fella retweeted
Qwen3.8-27B ships with reasoning_effort on xhigh. Turns out that's the one setting you don't want. Ran all four levels through the same config, 1083 tasks each. low and medium tie. xhigh burns 3x the tokens of low, loops 10x more often, and scores no better for it. I'd just run low effort. Config below. UD-IQ2_XXS since that's what fits 12GB, but I feel like higher quants will land in the same place.
48
40
18
576
54,370
Allen πŸ‡ΊπŸ‡¦πŸ‡ΊπŸ‡ΈπŸ‡¨πŸ‡¦πŸ‡ͺπŸ‡ΊπŸ‡¨πŸ‡­ Fella retweeted
free speed for every 24GB gpu owner, found overnight and open sourced before sunrise. context if you're new here: qwen 3.8 27b dense dropped last night, the strongest small model in the world right now, and i published day one benchmarks showing 25-31 tok/s raw on an rtx 3090. i also noticed something in the server logs, unused tensors. qwen trained a multi token prediction head into the model, a little draft brain that predicts tokens ahead so the big model verifies instead of generating every one. at 2am i posted that i was attempting to build the mtp draft file myself. here is what i found: there is nothing to build. the head already ships inside the gguf you downloaded, llama.cpp has supported it since july, and one flag connects them. measured properly, paired a/b, same card, same file: the rtx 3090 goes 31 to 41.3 tok/s, +33%. my 5090 laptop goes 36.7 to 50.9, +39%. acceptance holds around 78%. everything is in the repo, the flag, the exact launch commands, the tuning sweep, the probe script i measure with, and a community table waiting for your row: github.com/sudoingX/qwen38-m… this one is for you. flip the flag, then come tell me what your card did.
36
45
5
702
35,755
Allen πŸ‡ΊπŸ‡¦πŸ‡ΊπŸ‡ΈπŸ‡¨πŸ‡¦πŸ‡ͺπŸ‡ΊπŸ‡¨πŸ‡­ Fella retweeted
NEW!!! Hawaii’s state legislature has formally petitioned U.S. Congress and the White House to supply Ukraine with all necessary weaponry to fully defend itself and retake territory under international law, per a newly filed House memorial 1/2
43
433
13
2,174
40,515
The bubble is not in the AI. Competition will bring prices down to near the cost of inference. The real bubble is in the rent seekers who planed to base their trillion dollar valuations on obscene margins.
1
129
Allen πŸ‡ΊπŸ‡¦πŸ‡ΊπŸ‡ΈπŸ‡¨πŸ‡¦πŸ‡ͺπŸ‡ΊπŸ‡¨πŸ‡­ Fella retweeted
SpaceXAI's Grok 4.6 scores 61 on the Artificial Analysis Intelligence Index, joining the frontier in line with GPT-5.6 Sol, with standout agentic performance at lower cost Grok 4.6 gains 5 points over Grok 4.5 on the Intelligence Index just over one month after its release, or +23 points compared to Grok 4.3. This brings SpaceXAI back to the intelligence frontier alongside OpenAI, behind only Anthropic. Key takeaways: ➀ Grok 4.6 joins the frontier of the Artificial Analysis Intelligence Index: It scores 61, in line with GPT-5.6 Sol (max), behind Claude Opus 5 (max, 63) and Claude Fable 5 (max with fallback, 62), and just ahead of Kimi K3 ➀ Strong agentic performance: Grok 4.6 achieves a GDPval-AA v2 Elo of 1753, behind only Claude Opus 5 and with overlapping confidence intervals with Claude Fable 5 and Qwen3.8 Max. It scores 50.7% on 𝜏³-Banking, among the top two scores alongside Qwen3.8 Max (51.3%), and 88.4% on Terminal-Bench v2.1, in line with the leading models ➀ Frontier-level intelligence at lower cost: Headline pricing is unchanged from Grok 4.5 at $2/$6 per 1M input/output tokens, 60%+ below Claude Opus 5 ($5/$25) and GPT-5.6 Sol ($5/$30). It cost $0.84 per task, the same as Kimi K3 with slightly higher intelligence, placing it on the Intelligence vs. Cost per Task Pareto frontier ➀ Grok 4.6 sits at Fable 5-tier on AA-Briefcase, our private benchmark of long-horizon agentic knowledge work tasks, with an Elo of 1577 - behind the Claude Opus 5 family. It is notably turn-efficient, completing tasks in ~53 turns and ~0.5B input tokens on average vs. ~103 turns and ~2.0B input tokens for Claude Opus 5 (max) Other model details: ➀ Context window of 500k tokens (unchanged from Grok 4.5) ➀ Pricing of $2/$6 per 1M tokens of input/output; cache hits discounted to $0.5 per 1M tokens, an increase over Grok 4.5’s $0.3 per 1M tokens for cache hits Congratulations to @spacexai and @elonmusk on the release!
164
366
181
3,419
9,274,954
Allen πŸ‡ΊπŸ‡¦πŸ‡ΊπŸ‡ΈπŸ‡¨πŸ‡¦πŸ‡ͺπŸ‡ΊπŸ‡¨πŸ‡­ Fella retweeted
Introducing Grok 4.6. It delivers frontier intelligence and is a significant improvement over Grok 4.5 at the same price.
1,859
3,337
2,671
30,212
26,939,766
Allen πŸ‡ΊπŸ‡¦πŸ‡ΊπŸ‡ΈπŸ‡¨πŸ‡¦πŸ‡ͺπŸ‡ΊπŸ‡¨πŸ‡­ Fella retweeted
I've measured the exact SuperGrok & Cursor usage limits using Grok 4.5 Exactly 246,975,735 tokens account for 100% of SuperGrok's weekly limit with Grok 4.5 set to "Low". Compared to Cursor's $20 plan - SuperGrok provides 271,825,760 more tokens in total per month. In other words, SuperGrok has a 37.96% advantage over Cursor in terms of total token count. But SuperGrok costs $30, which is 50% more than Cursor. If we account for the price difference, Cursor would provide an equivalent of 1,074,115,770 tokens in total, which is 8.73% more than SuperGrok. Another plus of Cursor is that its limit is monthly rather than weekly, meaning you can have longer or more intensive sessions without hitting a weekly limit.
92
30
11
648
46,743
Allen πŸ‡ΊπŸ‡¦πŸ‡ΊπŸ‡ΈπŸ‡¨πŸ‡¦πŸ‡ͺπŸ‡ΊπŸ‡¨πŸ‡­ Fella retweeted
How to Surrender to a Drone? Ukrainian drones equipped with loudspeakers or dropping leaflets with surrender instructions instead of grenades have become a common sight on the battlefield. For Ukraine's Defense Forces, it does not matter how Russian soldiers who no longer wish to fight surrender. The important thing is that they lay down their arms and become part of the prisoner exchange pool. ☝️ As Russia runs out of recruits, it is bringing in foreigners from all over the world. Our colleagues @StratcomCentre have produced a video guide for them explaining how to surrender to a drone. It's very simple: there's no need to improvise – just let the drone operator know that you intend to surrender. ❗️Please share this video. It could literally save someone's life.
14
79
5
327
24,285
Allen πŸ‡ΊπŸ‡¦πŸ‡ΊπŸ‡ΈπŸ‡¨πŸ‡¦πŸ‡ͺπŸ‡ΊπŸ‡¨πŸ‡­ Fella retweeted
When the Roman empire began to crumble, it also retreated troops from their farther territories to their core. A similar decline can be spotted right now, with the American empire.
20
42
435
12,303