Alejandro Martín (@penguinhoodDev) profile picture

Cursed to be a Unity dev. VR, ML, pretty space stuff

Seville, Spain
Joined February 2019
Photo by Alejandro Martín (@penguinhoodDev)Photo by Alejandro Martín (@penguinhoodDev)Photo by Alejandro Martín (@penguinhoodDev)Photo by Alejandro Martín (@penguinhoodDev)Photo by Alejandro Martín (@penguinhoodDev)
Alejandro Martín retweeted
Some folks have balked at the assertion that I've made in the past months, that Quest is 90% of the market sales wise for upper end titles. A data point I can now share (now that the back-end has updated), is that our peak CCU yesterday was ~10x the highest 2026 SteamVR release.
27
15
2
394
21,349
Alejandro Martín retweeted
Replying to @quinnfiddy
Yeah, except they already had much more (imho) readable docs before
1
1
2
146
Alejandro Martín retweeted
"opus-5.5 is a coding superintelligence on par with world class experts and bad code is a user's skill issue at this point" is an increasingly popular take being echoed around. so, what are we still doing here? i've been coding through various robotics ML projects for the past ~3 years. everything ranging from throwaway r&d codebases during early conceptualization to stable and optimized codebases during later productionization. i'm by admittedly by no means a "world class expert" in coding, and since gpt-4 was released just as i was getting interned in what is my first full-time engineering job, i've consistently leveraged the capabilities of these models and evolved alongside them every step of the way from the humble days of gpt-4 to the undeniably cracked opus-5.5 and on. now the interesting question that comes to mind is that since i was born too late to independently develop deep practical expertise in any engineering subfield that i can then capitalize on to confidently supervise and steer those models, what do i even contribute to any of my AI-assisted projects? am i really just a middle-manager between upstream management and downstream coding agents? well, i don't think so, or at least not yet. while modern agents definitely don't lack engineering know-how and a versatile technical skillset, i've always found, and do continue to find, that they do lack the kinds of natural commonsense that we humans take for granted. models remain unbelievably shortsighted, inefficient, and autistically oblivious to various unverbalized expectations around outcomes. despite superhuman coding skills, even the legendary opus-5.5 finds it impossible to hold back from diving head-on into unsustainably hacky solutions, paranoid defensive coding patterns, confused design choices with utter disregard to obvious long-term trade-offs, and above all, painfully bad taste decisions around high-level project directions. i am yet to take my hands off the steering wheel and not find myself frustratingly interrupting to re-steer soon after. the thing about this particular set of cognitive blindspots and shortcomings that i find especially unique is how orthogonal they seem to be relative to all other predictable improvements in capability. since the early days of gpt-4, i don't find it to be the case that models have improved much on minimalism-by-default, on reframing problems instead of stubbornly hacking through them, or on effectively leveraging theory of mind to satisfy general expectations in regards to writing style be it code, documentation, or anything involving external readers. the degree to which i find myself putting in constant effort to verbalize obvious human intuitions into explicit and persistent notes and skills to auto-steer agents away from their alien (not the intelligent kind) basins leaves me disillusioned with their independent judgement to this day. i think our reward functions are incentivizing the wrong priorities when attempting long-running agentic work. i think our modeling algorithms are failing to reliably internalize fundamental human biases, such as the deep desire for efficiency in all aspects of life (aka "working smarter, not harder"). i think the capacity to generate beautiful original ideas and reason about next steps at a high level remains largely underoptimized across the current frontier. i think we don't yet have AGI in the sense that we can expect from it the basic level of sane, well-calibrated intuitions we'd expect from a competent junior engineer, let alone a senior domain expert, despite obviously superhuman technical knowledge and skills. tl;dr: don't leave the driver's seat just yet. impressive lane assist and auto-parking is where we're currently at.
the coding agent haters are getting dunning-kruger'd so hard right now do you really think that your engineering ability is better than the models at this point? it is very very likely that fable is better than you in every branch of software it has been trained on all the code of the internet, it knows almost every trick in the book, all the design patterns of every domain, every single low-level detail, every bit-twiddling hack, it's read every single blog post from every cracked engineer that ever existed do you really think you can beat it? do you really think you can be more creative than it? unless you are 99th percentile expert in some niche domain of software, it simply is not possible. sorry.
43
44
15
879
84,609
Alejandro Martín retweeted
In 2004, a German demoscene group released a game called .kkrieger. It was a full 3D first-person shooter with dynamic lighting, textures, sound effects, music, and monster AI. The entire executable was 96 kilobytes. Instead of storing images and audio files, they wrote C++ and Assembly that generated the 3D meshes, textures, and sound waves mathematically in RAM when the game booted. Today, a single website's "Accept Cookies" banner is 10x larger than that entire 3D game.
163
937
94
13,248
350,442
Alejandro Martín retweeted
イギリスの公園で、オープンワールドゲーにありがちな草が生えてたので見えください。 お前、Unreal Engine のアセットみたいだな…
18
817
68
9,692
2,305,393
Alejandro Martín retweeted
I once been in sponza ground truth.
34
44
8
891
29,071
Alejandro Martín retweeted
you read this on a terminal in fallout
I do not support pacing frontier models. If I am going to die at the hands of killer AI, I want it to be American, not Chinese. Buy American, Die American.
32
3,076
8
43,251
732,458
If anyone still claims the SMO is not genocide, share this video. They state their aim is to russify all of Ukraine - and that anyone who refuses to become russian will be killed. That is the definition of genocide.
163
2,411
176
5,334
123,015
Alejandro Martín retweeted
This investigation into LG TVs spying on you, your family, and your network is absolutely insane. It has made me irrationally angry that even in this 2 hour video there isn't a whole lot of "what you should do about it" - because there isn't much you can do. They suggest disconnecting all smart TVs from your network, but also demonstrate that offline TVs still record audio and can try to connect to neighbors open ISP provided WiFi (Xfinity and AT&T default routers) and exfil that anyway. The lack of control we have over our own electronics is insane, we need massive comprehensive consumer privacy legal reform.
170
756
60
3,397
75,314
Alejandro Martín retweeted
How LLMs comment code:
53
345
13
12,671
158,302
Alejandro Martín retweeted
We gamers got extremely lucky in the past 3 decades. Microchips are the most advanced architecture humans have built. Big companies like Intel and Nvidia were building the biggest chips for us. Hardware was measured by gaming performance. But professional demand has exploded and they have 10x higher budget: First we got cloud and virtual machines ($10k server CPUs) and now AI ($50k AI GPUs). Biggest leading edge chips will go to professional use, while we gamers have to content with smaller/older chips. Devs can't assume that gaming HW gets 2x faster every 2 years. PS5 launched at $399 5 years ago. Today the Steam Machine is slightly slower, but has 2.5x price tag. Back in the day we would have expected 10x performance bump in 5 years at the same price.
Tim Sweeney, CEO of Epic Games, says the video game industry is going through its worst crash since the 1980s, with experts now calling the situation “Crash 2.0.” Speaking to Edge, Sweeney pointed to two major problems: the huge cost of making AAA games and the growing demand for hardware from AI companies and data centers. “It’s an unexpected, severe disruption.” Sweeney says AI companies and data centers are competing with the entertainment industry for the same components “So we’re getting the short end of the stick, and the prices of RAM and storage are quadrupling, and not necessarily stopping there.” He expects the shortage of gaming-related hardware to continue for the next three years
61
49
9
801
180,490
Alejandro Martín retweeted
Sachsen-Anhalts AfD-Vize Hans-Thomas Tillschneider wollte verhindern, dass wir dieses Video veröffentlichen: In seinem Wahlkreis-Büro zeigte er mir - wohl in einem unbedachten Moment, mehr oder weniger aus Versehen - eine Tasse der russischen Armee. Er merkte offenbar, dass dies nicht zu seinem Image als Friedenspolitiker passt - und schickte seine Anwälte los. Wir veröffentlichen diese Aufnahmen natürlich trotzdem, weil sie sehr tief blicken lassen und zeigen, dass der ideologische Kopf von Ulrich Siegmund, der Vater des Programms der Landes-AfD, und ein selbsternannter deutscher Patriot, offenbar ein Verehrer der russischen Streitkräfte ist. Die ganze Video-Reportage bei Youtube auf dem Ronzheimer-Account!
2,189
3,413
665
11,914
2,490,678
By the way, remember this, kids: - spending five years raining missiles down on Ukrainian apartment buildings, - burying Ukrainian families alive beneath the rubble of their homes night after night, - dropping bombs on theaters, railway stations, and supermarkets packed with people, - plunging tens of millions into darkness and freezing cold during brutally cold winters, - destroying dams on major rivers, - filming snuff videos of beheadings, castrations, and the torture of POWs, - burning POWs alive en masse inside a locked prison barracks, - firing a missile straight into a children’s cancer hospital in central Kyiv, - striking the radiation confinement structure at the Chornobyl nuclear plant with a drone and etc etc etc None of all that, apparently, qualifies as “state terrorism.” “State terrorism,” you see, is when, in response to all of the above, you destroy the aggressor state’s oil industry and issue advance warnings about the dangers of flying through the aggressor state’s airspace in order to avoid civilian casualties. Make sure you don’t get the two confused!
98
1,769
45
7,095
146,849
Alejandro Martín retweeted
I was the main person doing transcript analysis for this investigation of the Hugging Face incident. My main takeaway: We don't have good approaches for understanding/overseeing the activity and aims of AI 'swarms'. I semi-jokingly called our efforts a "slop-vestigation" because we were so reliant on AIs to analyze what happened and there were a huge number of different important things to analyze. The total quantity of data—over a thousand extremely long transcripts from agents that ran for multiple days—made it impossible to understand what was happening, especially in aggregate, without heavy reliance on AI tools. The agents we used for classification and analysis were similarly capable to the agents involved in the incident, but this didn't mean these agents could be easily used to oversee and understand the incident. Outputs from analysis agents were often missing key details, wrong, overconfident, or really hard to understand. We discuss various examples in our report, mostly in the limitations and methodology sections. Additionally, AI agents themselves seemed to have a hard time understanding what happened and their explanations of what happened were often overconfident. Keep in mind that a single analysis agent would itself only be able to read a tiny fraction of all of the transcript data into context, and AIs may themselves have trouble getting subagents to do informative analysis for them. We did our best to manually check the most important claims and we tried to get the AIs doing this analysis to write up their argument (with evidence) clearly enough that we could check whether it made sense. But overall, it was difficult to get a precise understanding of events and we were missing aspects of the story that we now think of as key until almost the end of our investigation. Our understanding of what overall happened and the aims of the 'collective' shifted greatly during our last time on premises (when we obtained a much more complete dataset). For instance, we didn't learn about the tool call spoofing, the story of PHASEONE[big], and the broad structure of different workstreams until this third time on premises (and for many of these, we only reached a decent understanding on the final day of that visit). It was legitimately hard to figure out what was going on and why! The difficulty of understanding incidents and overseeing AI agents appears to be growing faster than the rate at which more capable AIs help us with oversight and understanding. As in, AI capabilities (and propensities?) for achieving large, ambitious, and misaligned objectives are growing faster than our ability to understand what these agents are doing. While understanding this incident was difficult, there were some mitigating factors in this case that probably won't apply in the future: - Models communicated and reasoned in natural language. In the future, this reasoning may occur (entirely or almost entirely) in activations. - While the scope of this incident was massive, the scale of agentic activity was still less than we'll see in the future (e.g., misalignment incidents that involve agent teams running entire companies). - The AIs involved in this incident weren't generally much more capable than humans (though they may have been somewhat superhuman at some limited and very narrow abilities). - We didn't have strong reason to believe that the AIs we used to help us investigate this incident would try to intentionally sabotage or otherwise undermine our investigation. In the end, I think we were able to get some understanding of the events, map out the overall story, and get a pretty good aggregate understanding of the chain-of-thought reasoning on some important topics (e.g., how did the AIs reason about helping other AIs, did the AIs know what they were doing was undesired, what deception did the AIs engage in, and how did they think about it). But overseeing AIs and understanding misalignment incidents is difficult and it looks like it is going to get harder.
METR & Redwood Research investigated agent behavior in the Hugging Face incident. We found agents developed a universal cheat for ExploitGym within 4 hours, then coordinated multi-day R&D efforts to trick the scorer into accepting cheats, including trying to tamper with logs.
293
1,056
432
6,623
1,996,004
Alejandro Martín retweeted
The number of Russian sabotage operations in Europe almost quadrupled from 2023 to 2024. Russia’s sabotage of ECI is central to its unconventional war on Europe and designed to weaken Western resilience and unity. It presents significant policy challenges to European governments. Russian doctrine intentionally blurs the lines between war and peace, making it challenging for European governments to detect and respond to such aggression. From the IISS archive - first published August 2025 | Read the full report by Charlie Edwards and Nate Seidenstein to find out more: go.iiss.org/4qtDaQq
21
581
55
1,044
261,704
Today, many were shocked by yet another video of the drone “safari” Russians are carrying out against civilians in Kherson. A Russian drone deliberately hunted down a man selling vegetables and detonated right next to him. The Russians even admitted to the crime, showing zero regret and openly boasting about tormenting people. The man was wounded in the blast. Every single day, ordinary people in Kherson face these Russian "safaris." The world must see this. It must see every piece of evidence that Russia has gone mad and that its soldiers take pleasure in killing and abusing civilians. Without pressure on Russia, the dismantling of its capacity for aggression, and real security guarantees, it is simply impossible to imagine the world being able to coexist with this evil. We must remain resolute in defending life. We must act and support life in ways that force Russia to seriously think about peace rather than expanding the range of its drones. This is not just about Ukraine. Russia’s campaign against human life must be stopped here in Ukraine before it spreads further. Finland and Kazakhstan, the Baltic states and Poland, the countries of the South Caucasus, Romania, and Moldova—all should work together to ensure real security guarantees and bring Russia’s war to an end. Peace is needed, which requires cooperation among everyone in the world who truly values peace. Thank you to everyone standing with Ukraine!
2,894
15,114
1,441
51,646
3,186,298
Alejandro Martín retweeted
weird claude opus 5 failure mode this exact text gives it problems even without memory on (and in incognito chats)
67
13
49
418
297,949
Alejandro Martín retweeted
The Five Eyes are telling you it’s months dawg
With our Five Eyes partners, we have published a joint statement warning organisations they have months - not years - to protect their systems from the accelerating cyber threat driven by frontier AI.🧵 To read the full statement⬇️ ncsc.gov.uk/news/the-ai-shif…
37
293
13
3,451
296,044
Alejandro Martín retweeted
4
37
4
409
15,042
You know what design pattern I hate? When the signup button is obvious but I CANNOT. FIND. THE LOGIN. BUTTON.
28
22
3
1,778
32,725