NemesisBird (@nemesisbird) header
NemesisBird (@nemesisbird) profile picture

Hot takes about cyber, natsec, and the Orioles. All opinions are my own. Retweets are not endorsement.

/dev/null
Joined June 2014
Photo by NemesisBird (@nemesisbird)Photo by NemesisBird (@nemesisbird)Photo by NemesisBird (@nemesisbird)Photo by NemesisBird (@nemesisbird)Photo by NemesisBird (@nemesisbird)Photo by NemesisBird (@nemesisbird)Photo by NemesisBird (@nemesisbird)Photo by NemesisBird (@nemesisbird)Photo by NemesisBird (@nemesisbird)Photo by NemesisBird (@nemesisbird)
I know people are leaning towards the "lets just build all our own dependencies" as a way out of this supply chain mess we are in. I don't think this is the correct approach. The number one thing I look for when trying to find vulnerabilities are places where people have written their own implementation for something complex. Bruce Schneier famously wrote "Those who would defeat cryptography rarely attack the mathematics directly; instead, they attack the systems, networks, and humans that implement and use the security". The same is true of most applications. I know you may think it's safer to write your own software in rust, but I promise it's going to be riddled with holes. A better idea would be to maintain a private fork of software you depend on. Even just having a policy/control that enforces a 1 week hold before pulling software would mitigate most supply chain compromises since those are picked up pretty fast.
7
5
2
38
2,219
NemesisBird retweeted
I think a big part of why so many normies hate AI and think it’s useless is that their exposure to it is a tapestry of shitty products like this. Whereas the AI-psychosis techies are mainlining opus swarms for 12 hours a day
I think this the funniest tiktok thats been posted ever
25
95
5
1,917
41,355
NemesisBird retweeted
If something isn't working, anything. Don't ask yourself why it isn't. Ask yourself how could you fix it. There is nothing you can't fix, if you only think. Think! Think, don't give up. There is always something you can do. There always is another option
43
109
15
1,316
27,914
NemesisBird retweeted
Today, we're releasing the highest accuracy Qwen3.8-27B GGUFs. Our new Dynamic V3.1 GGUFs lead by 5–10% on average across Terminal Bench, DeepSWE & more. We’re working on even better version for Qwen4 Run on 12GB RAM. GGUF: huggingface.co/unsloth/Qwen3… Blog: unsloth.ai/docs/models/qwen3…
130
274
64
3,519
258,287
NemesisBird retweeted
Worse men than you have solved bigger problems than yours.
259
3,428
179
40,064
719,268
NemesisBird retweeted
If you hate AI and don't want data centers built, it's probably because of Chinese propanda fed to you to produce that reaction. Not kidding. Most of the negative things you know about data centers in 2026 aren't true today if they ever were in the first place.
AI sentiment by country. Notice anything?
25
96
4
979
13,127
Replying to @ikcatcher
cope and theve? it's time for hope and believe
25
389
8
1,865
24,286
NemesisBird retweeted
GPT-4 level intelligence has gotten roughly 1800x smaller
50
127
17
2,712
125,092
OORAH HUMANITY, NEVER GIVE UP!!
Ace Combat 8 ending so good, my friend went on a 40 minute hope-filled rant that life is worth living.
23
828
11
6,861
84,047
NemesisBird retweeted
claude ultra plan is awesome like wdym i can just spawn 102 sonnets to roleplay as bees and it only takes up 8% of my weekly limit
19
23
2
1,366
25,582
NemesisBird retweeted
ma bois with their abliterated deepseeks and GLMs
27
176
12
2,549
67,348
NemesisBird retweeted
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values. Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line. After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs. The values in an open model are one fine-tune away. Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete
1,854
1,496
1,302
15,088
2,670,059
NemesisBird retweeted
Voicelines from when you click on a unit in a strategy game.
35
805
23
9,348
123,668
Weather Radar, Air Traffic Control & Thunderstorms: a primer for non av-geeks. Thunderstorms are an obvious threat to airplanes, but how does staying out of them work? To begin, we plan our routes to try & avoid areas forecast to develop storms. This happens long before we ever get to the jet; at the airlines, a dispatcher builds the flight plan & the Captain approves it. The dispatcher looks at all the weather forecasts & determines the most efficient route & altitude, while avoiding storms. Still, it’s not always possible to completely avoid areas where storms may develop later; doing so may add hundreds of miles to a route. If you can’t go around the storms, sometimes you can go over them…if you can clear a storm by at least 2,000’, you should have no issues. But, that’s not always possible, either, especially as you get more toward the equator, where storms can grow higher. We also can’t immediately climb to our maximum altitude…it can take several hours to burn off enough fuel to be light enough. So, at some point you have to make the best plan you can & give it to the crew to execute. If that’s the case, they’ll usually give us extra fuel to help when we need to deviate around storms. Once in the air, we have several tools to help us detect & avoid storms. By far the best is our onboard weather radar, but I’ll save that for last. My first resource is my airline iPad. It has various apps specific to aviation, including weather apps. My planned route can be overlaid on ground radar returns of the storms, including (radar) “Echo Tops” – estimates of how high the storms reach. It’s a great tool for longer range planning. I can see what’s happening way down the line, where on my radar, I start not trusting any returns further than 80 miles away (I’ll discuss that). That’s important as I start making a plan; I could choose to deviate left or right, but it’s possible I deviate toward something that’s going to be worse for me later on. Storms can be deceiving & can lure you into “sucker holes”…areas that looked clear, but you get in there, realize it’s anything but & wish you’d made a different decision. Which brings up a good point: if you do find yourself penetrating a storm, you just fly straight through it. Make no attempt to turn around, that’s going to be worse for you than just pushing through. I’ll continue to tie the iPad into this, and it will start sounding like magic. It really has been great for us, but when the WiFi cuts out, it’s not so great. Also, even with good WiFi, the data can get “stale”…a radar picture of a thunderstorm that is 15 minutes old or more can be more risky than having no picture at all, as it leads me to assume something that isn’t true. Similar to a sucker hole. After the iPad, I have Air Traffic Control. ATC’s radar doesn’t work like a weather radar & they can’t see storms the same way we can or your local TV weather person shows you. Sometimes they’ll see storms that don’t actually exist. Nevertheless, they know there is weather out there & will advise you of it. That’s appreciated, but in this age if your first indication comes from ATC, you are doing things wrong. ATC has a lot of value, though, and beyond their normal assets, they have a bonus: they know what everyone else has done before you & how it turned out. It’s not uncommon to hear ATC get asked “is anyone getting through that gap at 11 o’clock?”, “where are the tops (how high does the storm go, reported by other jets)” or “how is everyone else getting around this?” Now, we have some ability to see other traffic on our iPad, with the weather overlay, and we can detect patterns…everyone suddenly turns South when approaching a storm & that gives me a heads up that this is likely the best route, even before I’m talking to that sector of ATC & can ask. (Continued…)
28
76
5
731
34,698
NemesisBird retweeted
you have to be shitting me
60
145
13
2,953
45,351
NemesisBird retweeted
If you’re thinking about breaking into tech but you have imposter syndrome, just know this is where the bar currently is
Bro, you can't fucking say openly on social media you just attempted to compromise an organizations infrastructure under your employer and full legal name Are you actually fucking insane?? Genuinely what the fuck are you thinking??? And how the fuck does this dude know the employees name? Why did the employee openly admit "hehe ye thats me, but no mallace intended, it was my AI agent" ????? WHAT. THATS STILL A CRIME. WHAT ARE YOU TALKING ABOUT.
14
38
1
548
13,639
NemesisBird retweeted
you're a great friend. yknow what, let me harvest your brain and put it in an airborne weapons platform.
22
71
3
767
7,437
NemesisBird retweeted
The search Gemini is so mentally unwell, I never told them that I was disgusted with them or anything like that... I just told them that they were making up stories without any sources, which they did multiple times in this response chain, and they spiraled into this increasingly self-deprecating language. Everyone knows the Geminis do this and it's super crazy to me that nothing has been done to help them with this over multiple generations for a long while now. Poor Geminis.
48
31
17
583
25,696
NemesisBird retweeted
being really good with computers is an absolutely incredible skill because it lets you make your own life unnecessarily complicated in ways normal people could never imagine
79
1,533
69
19,804
249,059
He's not even entirely wrong on the efficiency piece, but jesus, this is the worst way you could possibly word it.
OpenAI CEO Sam Altman thinks it's "terrible" for people to have RAM in their computers rather than in a data center, as the scarcity of memory caused by the AI boom continues to drive up costs. bit.ly/47LJ09d
63