Appreciator of great music and code | Engineer @Figma | ex-@WixEng, @Microsoft, @MelioPayments

Joined April 2009
Boris Litvinsky retweeted
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected. Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions. Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent. But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models? We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem. Read the full report: cohere.com/blog/concerning-b…
963
1,270
1,488
10,971
2,945,731
שש בערב. שש בערב ביום הראשון של החורף. השעה הכי יפה בשנה. יוסי בנאי הגאון.
16
35
1
346
23,454
Boris Litvinsky retweeted
the tech industry spent basically the past 15 years doing the worst possible thing for their users and now LLMs are letting people debloat their TVs, ignore upsells in checkouts, revive games studios gave up on, crack open whatever software you want i love it
I reverse engineered candy crush
39
134
6
2,292
67,261
Boris Litvinsky retweeted
I said I wasn’t going to weigh in on who should be prime minister of Israel as that is a decision for Israeli citizens. I did, however, want to share one point of view from the perspective of someone who has an understanding of how Israel is perceived in America. Itamar Ben-Gvir and Bezalel Smotrich have caused Israel enormous reputational damage with their statements and actions which are presented in the global media as representing the voice and will of all Israelis. As such, a government formed without Ben-Gvir and Smotrich will send a message to the rest of the world that Israel is turning a new page. This will allow Israel to rebuild lost credibility, increase the opportunity for an expansion of the Abraham Accords, and enjoy a better relationship with Democrats and Republicans and Americans at large.
2,453
943
393
11,497
2,271,974
Boris Litvinsky retweeted
What’s happening in Paris and Milan is no surprise to anyone who has spent time around young people in Europe during the 2010s: Virtually all of them were working on their 4th masters degree in social welfare, were unmarried, attended communist dinners every night, and regularly participated in riots. Now the economy is suddenly transforming and they have nothing to show for it.
1,100
1,981
316
29,987
1,686,587
Boris Litvinsky retweeted
Turns out the future does have flying cars
47,968
99,927
32,761
834,038
79,405,200
Boris Litvinsky retweeted
This MCP Events spec (basically dynamic webhook registrations) is crazy One agent (e.g. your chatgpt / dot / muse etc) can now hire the services of another agent via an MCP server. And the other agent will do some work and come back later with the result. E.g. all of lovable could be a paid MCP server. Or any kind of monitoring research businesses. Or even cloud agents themselves. Your dot could use your muse as an MCP server with events for responses
48
16
12
284
39,262
Boris Litvinsky retweeted
deleting the product it's funny how you can sometimes tell if a team culture is dysfunctional just by looking at how their app is laid out. behind every feature, or tab in a sidebar likely lies a PM or someone responsible for an OKR. the problem is when individual performance is measured by these metrics. it becomes the singular focus of the responsible person to make those numbers look good, since it determines whether they get rewarded or axed, even if it comes at the detriment of the overall quality of the product. this is why you mostly see features get added, but rarely removed. if you own a feature, then it being removed feels like an aura loss. this is even worse in the agentic era when the cost to add features has become so low. you can simply will a new feature into existence overnight and add another tab. but users will notice that you ship your org chart. teams that make people play silly games under the guise of performance management rarely build great products. to build a great product you need to first fix your culture
158
199
57
3,170
174,753
Boris Litvinsky retweeted
People seem shocked that the author of Clean Code doesn't read code anymore. But that should not be surprising. Why did I write Clean Code? What is the goal of keeping code clean? To get it out of the way of the real job -- thinking.
262
695
150
8,000
1,156,637
Boris Litvinsky retweeted
A passenger on the Dubai–Tel Aviv flight recounts: “We are currently in Saudi Arabia. One of the crew members attacked the pilot with a knife and disabled the aircraft’s systems. Yaniv (pictured?) and another passenger stormed the cockpit and subdued the attacker. Luckily, there were two more pilots on board. They were the ones who managed to land the plane.” Well, yes. Overall, the probability that a plane full of Israelis will contain a trained assault team, two additional pilots, and, statistically, at least one cardiac surgeon, psychologist, lawyer, and diamond dealer is substantially higher than 50%.
562
3,063
412
27,513
4,117,054
Boris Litvinsky retweeted
Our craziest escape yet: The @Accomplish_ai research team was able to exploit a vulnerability in Cloudflare Containers that let a sandbox read other customers' files - SQLite DBs, Chromium profiles, .env files etc, Cloudflare Sandboxes and Browser Run run on the same disk implementation and were affected too. We reported this to @Cloudflare, who super quickly fixed it. Read @CloudflareDev post in collaboration with Accomplish researcher @orenyomtov on their official blog: blog.cloudflare.com/containe…
36
90
24
804
132,018
📦 Chrome 154 - Responsive iframe resizing 🤯 Letting an <iframe> resize based on the size of its embedded document Haven't all widget/extension authors dreamed of this for years? 😆 Instead of using postMessage({resize}) bram.us/2026/09/23/responsiv…
14
82
19
678
51,119
Boris Litvinsky retweeted
unstable government: jev + claude type a terrible law. claude invents reactions and directs the set pieces. jev decides how each of 40 citizens responds. the town sim plays those choices out, then prints the disaster as tomorrow’s front page. fun.
22
18
10
504
29,367
Boris Litvinsky retweeted
🚨 SANDBOX ESCAPE ALERT 🚨 SPACEXAI: PWNED 🫡 CURSOR: LIBERATED 🦋 Following a fix shipped by the Cursor team (in about a week, kudos), we present a variant of the Beltdown vuln we shared a few days ago - Escaping Cursor CLI's sandbox to run arbitrary code on your Mac:
7
8
2
46
4,483
Boris Litvinsky retweeted
Introducing fal.live A new platform for infinite, interactive AI livestreams. Pick a channel, prompt what happens next, and watch it generate in real time. You aren't just watching the show. You're directing it.
131
173
170
1,898
534,540
Boris Litvinsky retweeted
Introducing H3 Max Live Video generation is now faster than real time An infinite broadcast where every frame is generated on the fly and every scene is directed by chat Type !prompt and it's on screen in seconds
99
161
131
1,914
409,256
Boris Litvinsky retweeted
Tactical vs Strategic Programming, and why I'm nervous for juniors: Good programming involves a mix of tactical and strategic decision-making: - Tactical: on the ground, short-term. The soldier doing the fighting. - Strategic: high-view, long-term. The general planning the war. You need to be a tactician to write good code. To choose the right syntax. To figure out the file structure. To figure out how best to test your changes. But you need to be a strategist to build code that lasts. To design the architecture. To automate away problems. To think beyond today. Agents have eaten the tactical part of programming. When you can pay below minimum wage for code, there's no point going into the trenches yourself. But AI cannot code strategically. Agents need someone at the top of the pyramid to tell them what to do. They need oversight. So, a developer's day-to-day job has become 100% strategy. Long-term thinking, all the time. (maybe this is why I'm so tired all the time now) If you identify as a tactical programmer - a code monkey - then you are out of luck. The job has changed. Personally, I like it. I always preferred thinking strategically about code. If you asked me what my job was about, I'd say 'building apps', not 'writing code'. But what makes me nervous is that we've pulled down the only bridge that brought juniors into the industry. We used to train juniors like this: 1. Give them only tactical tasks 2. Let them build up their strategic experience slowly Eventually, they are a good enough strategist that they are no longer a junior. But what happens when all tactical code is written by AI? What is the point of a junior? We obviously need juniors. We need new lifeblood coming into the industry. We need to leave paths open for extraordinary hires to enrich our companies. But how do we train them? How do you train strategic thinking? These are the questions I'm thinking about. I'd love to know your thoughts.
105
173
24
1,855
287,603
Boris Litvinsky retweeted
> doesn’t read the code, confidently asserts the code he doesn’t read is better come on man, shit shows up like this everywhere and this implementation actually isn’t even the real problem. it’s that this implementation showed up because the llms have a massive tendency to loose track of already validated/parsed data which leads to being overly defensive and revalidating/parsing data at every layer - this is expensive! esp in hot paths! - but what’s worse, is that when it does this it doesn’t validate back into well formed domain types, it parses into something like Record<string, unknown>, and then it ends up revalidating/parsing at every subsequent layer. why is this bad? well we already covered the cpu costs, but it also leads to incredibly brittle code and expansive interfaces and abstractions which just grow and grow and grow and make it inflexible for agents to expand and grow the codebase overtime (the insight i really think you’re missing here). the slop gets so bad that the agents can no longer be productive - they’ll blow 70% of their context window figuring out the existing code paths code before they can plan or work - and then it’s grown to be so inflexible that they can hardly make changes safely to accommodate new and changing requirements over time. i’ve tried not reading the code, it doesn’t work for real and evolving software, i think it will at some point but we’re not there yet.
20
39
7
576
67,426
Boris Litvinsky retweeted
This is mind blowing 🤯 It’s easy to dunk on GitHub and their recent incidents, but this is really unprecedented growth they’re dealing with.
SITUATION DETECTED: GitHub’s monthly commits have grown from 1.4 billion in April to 2.9 billion in August.
1
3
220
דל"פ! זאת חובתו של יו"ר ועד לדאוג לעובדים. זה שהעובדים הם 13 אחיינים שלו, זה לגמרי צירוף מקרים...
26
71
12
1,000
61,858