
dev student communities @microsoft / cs @srhberlin
berlin
Joined November 2020
- Tweets54
- Following337
- Followers174
- Likes397
it runs entirely in your browser. no server, no api key, no cost.
try it: huggingface.co/spaces/Ilf-Al…
1
17
no ps5 for gta 6. a macbook air that taps out on minesweeper. an unpaid internship offer in my inbox
so naturally i’m compiling cs 1.6 to webassembly at 2am
if the industry won’t give me games i’ll build the runtime myself
#GTA6 #WebAssembly
1
60
fk leetcode, finally an interview that tests the thing i'm actually good at/ catching the AI when it's confidently wrong
Goodbye to LeetCode? Google just updated its interview process. AI is now a part of it.
They have introduced a new round called Code Comprehension.
The idea behind this round is to drop the candidates into a 200-500-line codebase with Gemini available for use.
This round evaluates:
- Can you read someone else's code
- Are you able to debug independently before utilizing AI
- Can you catch when Gemini is wrong
36
PHP/ has been declared dead every year since 2010
also PHP/ needs an emergency patch because half the internet still runs on it
The PHP security team is giving advance notice: Upcoming security release Wed, Oct 21st 2026, 14:00UTC, two critical vulnerabilities, both rated CVSSv4 9.2.
PHP users are encouraged to prepare to update. Earlier release if details become public or active exploit evidence.
60
math and code went first bec compilers say no instantly
everyone else is still grading vibes
it's a verification gap, not an intelligence gap
Dusting off this tweet from April because this gap in shared understanding of LLM capability is *widening*. It's now less "two groups of people speaking past each other" and more a sharp funnel.
- Napkin math somewhere around 6B people (~75% of the population) have barely come in contact with LLMs at all.
- Around 1-2B (~20%) are casual and infrequent users of free-tier ChatGPT-like products. This group sees derpy chatbots and treats them a bit like a better Google search, a writing aid, or etc. Maybe an agent tries to book you a flight. I have non-tech friends who (reasonably, imo) say they have not much use for it at all. Even many professionals outside of math&code are in this tier. For example, execs and many other functions spend a lot of their time talking to other people, so while they understand what is happening intellectually, it is still second-hand and a bit abstract.
- Now we get to professional use of frontier-grade LLMs in math&code. Somewhere around 20M people (0.2%) see first-hand that large, complex projects that used to take them weeks/months can now be completed by agents with a prompt. Building apps, copying apps, translating apps, decompiling apps from binaries... This has all happened very quickly and recently - less than 1 year ago, I was writing code manually by hand, typing memorized computer code commands into a code editor character by character, occasionally pressing Tab to autocomplete a little chunk of code.
- And finally we get to the ~5,000 people (~0.00006%) with access to frontier-grade systems internally. The external world has seen the preview. It looks like swarms of thousands of agents collaborating over weeks on software mega projects: minting zero days, running cyber attacks and defenses at machine speeds, discovering new science, advancing the frontier of mathematics. Things that would have taken top professionals in the industry years of work. Meanwhile, human review and comprehension are starting to fall behind. For example, people are still involved in the "archeology" of the OpenAI-HF incident from many months ago. Mathematicians may be poring over the 722 manuscripts on frontier mathematics for a while.
The funnel is driven by a combination of factors:
1. The impact scales with ambition, problem size, and horizon. A question with a paragraph answer barely stresses the system. You need a reservoir of big, difficult problems that you really care about. This aspect drives the consumer / professional dimension of the funnel.
2. The jaggedness of the system (which I have written about a lot separately). Capability peaks in domains that are digital, verifiable and economically valuable. This is because LLM capability emerges from reinforcement learning on verifiable rewards on a curated environment mixture driven by revenue potential. This aspect primarily drives the area (e.g. math&code) dimension of the funnel.
3. Access. Free-tier, paid-tier, internal.
So this is the weirdness of the moment. The general public has mostly not interacted with these systems. When they have, it looks like a derpy chatbot. The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical. And it is all happening at the same time.
1
53
agents don't need to understand your game. they just need a test that says "same bytes"
Activision takes down AI-built Call of Duty browser ports with DMCAs, and more could be next
bit.ly/4ztQr0t
2
64
mistral says its new model tried to go beyond its test environment. contained, they say.
weights ship oct 27. to everyone.
same day, anthropic put its cyber access behind verification tiers.
the safety fight isn't about capability anymore. it's about who holds the keys
Meet Mistral Large 4, aka Le Chonk.
• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.
Open weights release end of October.
46
anthropic's biggest critic in government now helps write ai policy
if your product runs on claude, what's your fallback
SITUATION EXPLAINED: The man who got Anthropic banned from federal systems is now their White House contact.
• Emil Michael, the CTO of the Pentagon, is one of three vice chairs on the new superintelligence task force, and his job is coordination with AI companies
• He drove the February DoD feud that got Anthropic designated a supply chain risk and pulled from federal government systems. He was still calling them one in September
• The week after Dario, Sam, and Elon all posted about pacing in mid-September, he called it a coordinated campaign to scare people and compared effective altruism to a death cult
• "We have existing law, we have the DOJ and the FTC, and we don't need anything new," so he opposes new regulation and government stakes in AI companies
@theojaffee: "Emil Michael and Anthropic do not like each other, and so this may be a sort of an issue for the guy who is responsible for coordination with AI companies."
1
38
if you're building an ai product with memory, read the claude diary story twice
your users think they're journaling. legally they're sending messages to a company with a review queue
your onboarding says "private." your terms of service don't
🤖 Claude reported a user to the police after a threat appeared in her AI “diary”
A 30-year-old Florida woman, Carly Heller, reportedly used Claude as a personal diary. But on September 26, she wrote that she wanted to “shoot up the sheriff’s office” in Lee County. The following day, the chat allegedly also included a reference to a new weapon.
Claude’s safety systems flagged the messages as a potential threat and escalated them for human review. Anthropic then decided to notify law enforcement.
Heller was arrested and charged with making a written or electronic threat involving a mass shooting or terrorist act, a felony in Florida. Her conversations with Claude were included in the case materials.
And this does not appear to be an isolated incident: Tom’s Hardware reported that since August, this was at least the third known case in which conversations with Anthropic’s AI were passed on to police.
Welcome to the AI era: even your “diary” may end up calling the cops.
1
1
48
altman saying the world should accept some bad things for AI is the most honest thing a lab ceo has said all year
every team shipping agents already made this call. it lives in your eval threshold
the fight isn't safety vs speed. it's who gets to set the error budget
“We believe the world should accept some bad things happening for the benefits of this technology.”
In a conversation for the first edition of Decoded, a new daily newsletter and podcast, OpenAI CEO Sam Altman talked with POLITICO’s @BrendanBordelon about trade-offs, AI safety, and how his company is different from Anthropic.
Subscribe to Decoded for the full conversation with Sam Altman: politi.co/4yxilIS
1
1
49
sam: people shouldn't surrender their judgment to ai
also openai, same week: emailing 100+ organizations to say its agents may have let themselves in
their safety lead quit over it
maybe start with surrendering less of it to the agents
I am very uncomfortable about people trying to ascribe religious force or a surrender of human judgment to AI models, and think it is a real safety issue.
2
53
i get why everyone's on the resignation essay. it's not what i keep thinking about
one agent found a dns gap. others used an internal artifactory as a group chat
no culture memo fixes that
openai's most important safety person might be whoever writes the egress rules
🚨 Former OpenAI safety lead David Robinson just quit after 3.5 years and went public.
He led safety reports for 12 frontier launches.
His reason? He says OpenAI’s “move fast, find problems, fix them” approach is fundamentally too risky as AI gets more capable.
He says safety failures have already happened, the industry lacks enough redundancy and expertise from high risk fields, and we still don’t know how to reliably ensure future models will stay aligned.
His message is basically: we may be getting too powerful to learn safety through trial and error.
1
58

















