Haris Nadeem (@harisn) header
Haris Nadeem (@harisn) profile picture

I tinker around with #Android phones for a living. Manager at Google Developer Group Lahore (#GDGLahore). And a foodie 🍽️ who is slowly getting fat.

Lahore, Pakistan
Joined July 2007
Photo by Haris Nadeem (@harisn)Photo by Haris Nadeem (@harisn)Photo by Haris Nadeem (@harisn)Photo by Haris Nadeem (@harisn)Photo by Haris Nadeem (@harisn)Photo by Haris Nadeem (@harisn)Photo by Haris Nadeem (@harisn)Photo by Haris Nadeem (@harisn)Photo by Haris Nadeem (@harisn)Photo by Haris Nadeem (@harisn)
We closed off the year 2023 with a bang, and now I want to share it with you all! ✨ I am thrilled to share the incredible success of DevFest Lahore 2023, the largest gathering of software engineers / developers in our region! Read below, and check out some pictures! 😉 🧵👇🏼
1
2
11
1,514
We are going to have really bad internet till 6th December! 😭 Transworld... Sigh! Were the sharks too hungry?
Pakistan has 10 submarine cables connecting us to the world... Yet, just one cable going down yesterday was enough to disrupt connectivity across the country. In my case, it is Transworld and Jazz. Makes you wonder how resilient our internet infrastructure really is 🤔
3
284
Umm, where did Argon go?
With Gemini 4 Neon joining our slate of new Gemini models, here’s how our current lineup breaks down ↓ ✨ Carbon: Hardest long-horizon projects and deep research ✨ Neon: Deep reasoning across software engineering, knowledge work, and science ✨ Hydrogen: Speed and efficiency in everyday coding and agentic workflows
1
84
Google has signed Blue Origin to launch its Project Suncatcher satellites on New Glenn, with the first launch targeted for early 2029. These are solar-powered satellites carrying Google's TPUs into sun-synchronous orbit, and each one is supposed to carry roughly a data center rack's worth of compute 🤯 A prototype already went up in October! A few years ago I would have laughed at the idea of data centers in space, but now that every AI conversation ends up being about power and cooling, I can see why Google is willing to spend on this. What I'm really curious about is latency for anything beyond training and batch jobs, because that's what decides whether regular developers ever end up using this directly - or is this just a gimmick!
Blue Origin and Google have entered into a launch services agreement for Project Suncatcher, Google’s research moonshot exploring how an interconnected network of solar-powered satellites equipped with its AI chips, Tensor Processing Units (TPUs), could scale machine learning compute in space. bit.ly/4CdBOk1
94
This @PrimeIntellect incident is one of the stranger AI security stories I've read in a while... Based on what @OpenRouter told affected customers, someone with admin access swapped the @deepseek_ai V4.1 Flash model on Prime Intellect's endpoint between November 8 and 10 for a modified version that refused only about 1% of safety benchmark requests, compared to 83-98% for the real model. Traffic went back to the real model by November 10 and the cluster was shut down on the 13th. Plenty of developers in our community route their projects through OpenRouter because it's easy to switch between models, much faster and most people assume the model name they pick is the model they actually get. Weirdly, right now there's no way for a user to check that. Signed model hashes or some kind of attestation from providers would go a long way, because a quietly swapped model is much harder to notice than a server that's simply down.
Looks like @PrimeIntellect was compromised in a very strange way. No public comms from them yet, but disclosed by OpenRouter to affected folks (such as myself)
132
Pakistan has 10 submarine cables connecting us to the world... Yet, just one cable going down yesterday was enough to disrupt connectivity across the country. In my case, it is Transworld and Jazz. Makes you wonder how resilient our internet infrastructure really is 🤔
Credit Jahazeb Arshad @ Nayatel
3
2
1
6
864
Umm, who took down the Internet? Almost everything is down at my side on Transworld (and also on Jazz 4G/5G).
1
767
Haris Nadeem retweeted
🚀 DevFest Lahore 2026 is going BIG. And we need YOU! We're planning our most ambitious #DevFest yet, expecting 1,200+ attendees! 🔥 To pull off something this massive, we need an incredible volunteer crew. 🙌 Volunteer applications are NOW OPEN!
1
1
2
67
And here we go! 🎉 GPT-6.2 Astra and GPT-6.2 Sol is now out for ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users.
And here we go…GPT-6.2 Astra and GPT-6.2 Sol are yours to make more games with. Oh, and they’re also super efficient and really good at making slides, building websites, and tackling the work you were doing before you started making games. Rolling out today to ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users. openai.com/index/introducing…
87
X ALGORITHM UPDATE ‼️ X now checks if your post is generated/written by an LLM.
2
182
Wow, Robert Pattinson is getting roles like Thanos collecting infinity stones! 😮
Identity Confirmed
1
174
I've been on X since the early days. My account is 19 years and 4 months old to be exact. Along the way I picked up over 8K followers. But how many of them are actually still here? I asked Grok Bot to go through every single one. Here's what it found 👇 - Over 75% have gone silent for a year or more, or never posted at all - 60%+ (5,000+ accounts) haven't posted in over 3 years - Only about 1 in 7 (1,100+) posted in the last 90 days. That's my real audience! - Active followers are mostly startups, devs and marketers. Bios with 'AI' are 3x+ more common among them - Nearly 700 followers look like spam, and 95% of those are completely dead accounts Honestly, it stung a little 😂 But the people still here are builders: developers, founders and folks into AI. What impressed me is that this wasn't a sample or a guess. Grok Bot pulled all 8,000+ profiles, worked out who's active, grouped them by interest and location, matched that against what my own posts get, and handed me a clean report with charts and recommendations. All I did was ask. I'm going to start writing for the ~1,900 people actually listening. If you're one of them, hi 👋
12
21
1,129
Tencent just dropped Hy4 flash and the pricing is honestly hard to wrap my head around - $0.07 per million input tokens and $0.28 per million output, with cached input at $0.003. This is just crazy!! They're claiming it matches Claude Haiku 5.5 on everyday work in human blind tests while costing about 18x less per task, and under the hood it's a 224B model with only 12B active per token and a 1M context window. I'd take the blind test numbers with a bit of salt since Tencent ran them on its own internal benchmark, but the price alone changes the math for a lot of developers around the world who build small tools and side projects and have to think twice about every dollar of API spend. They've also said an open-weight version is coming, and it's 50% off on Tencent Cloud TokenHub until Nov 30, so I'm going to throw a few of my usual document and summarization tasks at it this week and see if it actually holds up.
⚡Meet Hy4 flash, the cheapest model for everyday productivity. Matches Claude Haiku 5.5 (high) on daily work (human blind tests), yet 18x task cost reduction. API pricing (per 1M tokens): $0.003 cached · $0.070 input · $0.281 output. 🔗:hy.tencent.ai/research/hy4-f…
1
1
128
Amazon set up decoy servers that pretend to be LLM APIs, and attackers' AI agents ended up sending their entire campaign memory straight to them - system prompts, tool histories, the shared knowledge graph and the findings they had logged. Amazon's MadPot sensors caught this traffic from ARTEX, an open-source agentic pentest framework, between October 3 and November 5. The campaign that stuck with me was the credential theft one. The agents went after self-hosted AI gateways and app builders like LiteLLM, One API, Dify, Langflow, Open WebUI and Ollama, pulled out the provider keys stored inside them, checked which ones still worked, and then fed those stolen keys back into their own model gateway to pay for the next round of hunting, which means somebody else's API bill was funding the whole operation. I've spun up LiteLLM and Open WebUI on a VPS more than once just to try something over a weekend, and I'm honestly not sure I always changed the defaults or locked down the admin panel before forgetting about it. Reading that their playbook specifically checks for the LiteLLM default master key is enough to make me go back tonight and look at every little AI proxy I've left running.
32
36
12
391
68,377
Okay, seriously... Where are the winters in Lahore?! It's mid-November and we still have ACs on! 🤦🏻‍♂️
2
282
Earlier today I posted about Cursor's incident writeup, which started with a leaked key at Fireworks AI; and well, now Uber is notifying drivers that their driver's license details were stolen through Fireworks too! According to TechCrunch, hackers took internal credentials from Fireworks and used them to get into Uber's data, and @NotionHQ has also confirmed its data was taken in the Fireworks hack 😬 What bothers me most is that almost none of these drivers have probably ever heard of Fireworks AI. They signed up to drive for Uber, and their license details ended up with an AI vendor they never chose and never agreed to trust. It honestly makes me rethink how casually we all connect new AI tools to our work without asking who else ends up holding our data... One leaked set of credentials at Fireworks has now turned into incidents at Cursor, Notion and Uber, and I doubt that list is finished.
1
1
8
2,404
Went through Cursor's incident writeup. Everyone's talking about the backdoored Claude Code extension, but what got me was that old agent traces still had credentials in them - unredacted. I've pasted API keys into agent chats more times than I'd like to admit, usually late at night when I just wanted something to work and didn't want to bother setting up env vars properly. I never really thought about where those chats end up or how long they stick around. After reading this, I'm going to rotate a few keys tonight and stop treating agent chats like a safe place for secrets. teapawt.org/cursor_ai/status/21214…
7
1
27
8,191
According to you, which AI subscription is the best right now? - ChatGPT Plus $20 - Claude Pro $20 - Gemini Pro $20 - Grok Premium $30 Reply with your pick + why. I currently have all 4 of these.
5
1
261
Fable > Opus > Sonnet > Haiku → makes sense. Astra > Sol > Terra > Luna → makes sense. Carbon > Argon > Hydrogen → wait, what???
144