Everyone is chasing TRUMP But did you know you can chat with Trump right inside 2077? We just one-shotted an app where you can enter your OpenAI API key and start chatting with Trump Try this right now: ✅ Go to 2077.codes/chat/86w6dqf6 ✅ Click on the "CODE" tab, then find the .env file ✅ Add your OpenAI API key under VITE_OPENAI_API_KEY ✅ Hit "Save" And start chatting with Trump! Your OWN keys. Your OWN app

We're still building :) New "Gaming Mode" is live, along with 4 new quests. Dropping next: a new in-game level, inventory management, and a marketplace!

Update on the DeepSeek integration We've been trying to get access to their API but they've switched it off, presumably due to too much demand (You can see it at DeepSeek.com and clicking on "API Platform" in the top right corner) We'll wait for the API access to commence the integration In the meantime, regular development will proceed as planned P.S.: On a personal note, I know we've underdelivered in the past 7-10 days. The destructive/distractive energy of Trump/Melania really hurt shipping momentum. But that's really no excuse and we resolve to do better

Some perspective on this entire DeepSeek situation: 1. DeepSeek is a legitimate lab and well-respected. The V3 was genuinely novel and performed above expectations 2. Reinforcement Learning (RL) used by DeepSeek is a genuinely novel improvement 3. DeepSeek was trained on OpenAI's o1 output. OpenAI remains the first to cross the "four minute mile", so to speak. DeepSeek just borrowed from their technique 4. The core innovation here is RL, which essentially means that DeepSeek improves itself through self-reinforcement. 5. No, this isn't a psyop. Yes, CCP is likely involved - impossible for any large org (DeepSeek parent org is a $8B AUM fund - okayish size for China). That's par for the course when dealing with Chinese businesses 6. The model is genuinely nice and has a lot of personality 7. No, this doesn't destroy Nvidia. And no, it doesn't destroy OpenAI - remember that these were the first to come up with this reasoning technique. That said, their path to profitability is much harder now that open source LLMs have caught up 8. No, Nvidia isn't rekt. The bulk of AI costs are in inference (i.e. delivering results) not so much in training. Inference at scale will keep demanding Nvidia GPUs As things stand, the hierarchy for models is still O1 Pro > DeepSeek R1. But given that O1 Pro costs $200/month and R1 costs $0, R1 gets an edge. This might change when OpenAI O3 launches. But until then, it's off to the races

The recent upsurge in the popularity of DeepSeek have made us recalibrate our approach to AI models. Previously, we were using Sonnet 3.5 as the default model for all generations. However, given DeepSeek's outsized performance, we feel that a one-model-fits-all approach doesn't work anymore. Consequently, we'll integrate multiple models into the code gen. You will have the option to choose from OpenAI o1-mini, DeepSeek R1, Claude Sonnet 3.5, and Gemini ETA within 24 hours

You can now connect your wallet to the 2077 dApp 😀 This opens up the next chapter of 2077 where you will be able to buy in-game items, characters, extra credits, priority generations, and more! AI x GameFi is now even closer. Dev note: We were originally rolling our own auth with Solana wallet adapters, but all new dApps get blacklisted by Blowfish by default. This throws up a horrible looking warning on Phantom. Getting whitelisted by Blowfish is trivial but maybe because of the mayhem over the last weekend, we couldn't get cleared fast enough. Hence the slight delay and the eventual switch to Privy

Apologies for the lack of interaction here the last two days. Like everyone, we were distracted by what transpired yesterday. The development continues