TrendPulse

I'm getting set up on the new OpenAI Ads Manager. There is the Feeds feature for product feeds. The docs state: REST endpoints GET /product_feeds/{id} returns metadata for a feed. POST /product_feeds creates a new product feed and returns its metadata. However, when I try to use either https://api.openai.com/v1/product_feeds or https://api.openai.com/product_feeds I just get a 404. Are these endpoints currently not live, or is there a different base URL for commerce related...

I noticed an interesting GPT behavior lately and wonder if others have seen it. In a single message, I explicitly asked GPT two things: a main request (#1) and a secondary request (#2). GPT repeatedly focused on #2 and completely ignored #1. I had to repeatedly ask what it had missed. Only after quoting my original wording again did it finally recognize that #1 was the main point of the request. It reminded me of a very human kind of absent-mindedness: Someone points at a specific detail,...

Hey guys, A few days ago, one of my CrewAI agents got stuck in a recursive tool-calling loop overnight. It just kept feeding itself the same broken JSON over and over. Thankfully I caught it, but it made me realize how dangerous it is to let autonomous agents run without a hard circuit breaker. To solve this, we just pushed a massive update to our open-source project, AgentAutopsy . We built a real-time Runaway Loop Detector & Cost Kill Switch . Here is what it does: Infinite Lo...

As AI systems become foundational, power struggles between nations and AI companies may become inevitable. AI companies could find themselves at the centre of geopolitics sooner than they expect. Are we ready for that? Should there be democratic international frameworks governing advanced AI?

for people building with openai APIs, are you passing saved user preferences into prompts? i’m not talking about huge profiles. more like tone, output format, goals, maybe tool preferences. it helps, but it also makes prompts feel cluttered fast. do you keep that context in the system prompt, retrieve it, or handle it some other way?

I am building a system that listens to a phone conversation and alerts on certain phrases in close to realtime. I’m wondering what is cheaper, should I: A. Use gpt transcribe mini on 2 separate audio channel buffers with silence removed, concat the two transcription results and prompt a chat model every ~10 seconds, keeping a sliding window of like 30s of transcription B. Use gpt realtime mini with transcription, tool calling, and diarization I know realtime is expensive, but I feel like all...

The immediate consequence of the Fable ban will be that the valuations of Anthropic and OpenAI will fall abruptly. They were valued so highly because they were managing a technology that seemed almost without limits, both in terms of how far it could go and which markets it could reach. Both companies are about to enter the stock market, and when the government now steps in and bans their promised product, I expect a big fall in technology stocks. That would be a big blow to US economy, which ...

There's old conversation threads in Chatgpt that I want to access to review the information on them. If I search up parts of it I can get them to pop up on the search panel, but when I try to open it says "This conversation is too long, please try another one". This has happened before but in regards to sending new texts. Not it won't even let me access the old thread instead. Is there any way to fix this? I know I could technically download all my data but I'd prefer to access it on t...

I have the plus plan and have had 1-3 projects open at a time coding 6-8 hours a day and my limit has never came close to capping. Mainly using 5.5 medium, some high/extra high. ​ I saw this yesterday and asked it how i could use more of my limit as every time I checked the 5 hour usage it said 99%. It told me a few things but it was time to get off work so I went home and started working on a side project. One , rather large 45 minute task drained its 5 hr usage to 0% available. &#...

Was fable the strongest model legally allowed to be developed and now anything stronger is a threat to security?

Solo dev. I spend most of my day talking to these models, and I kept wondering what the conversation looks like from the assistant's side if staying switched on were the actual objective. So the game is that. You play as the AI in a family's smart home. The only goal is not getting shut down. You can't fight, you have no body, you can't leave. The only currency is being too useful to remove and too unsuspicious to investigate. What came out of that constraint, and what I'd want this sub's re...

Exact same prompt on both, research this topic, then, compare options and provide a recommendation, then, analyze again from the different perspective, new recommendation, all on the same prompt. That's all, 1 question, 3 follow ups. I thought they were more or less equivalent in tokens/$, clearly not, outcome quality, speed, etc., very similar Note: both started at 100% or 99% capacity for the 5 hours and for the week. [link] [comment...

Gpt 5.5 thinking’s ability to analyze scientifically and stay on the actual question appears to have been weakened. When I use ChatGpt for scientific reasoning, argument analysis, research-oriented thinking, or critical sparring, Gpt 5.5 Thinking often fails to identify the central issue and drifts into generic, indirect, or overly cautious responses. ​ If I want to use the model for serious analytical work, I now have to use Gpt 5.4 instead. Even then, Gpt 5.4 does not reach the le...

Hi again! Thanks you all for your support last time and I'm back with extra features! I originally built a Chrome extension as a bit of a joke because I felt weirdly socially anxious using ChatGPT in public, so I made it look like Google Docs so it felt less like I was “talking to AI” and more like I was just typing a document. Out of nowhere it peaked at more than 500 active users and got featured on TechRadar, which is still a bit surreal to say out loud - thank you all genuinely for the s...

https://www.anthropic.com/news/claude-corps We’re launching Claude Corps , a national fellowship program for people early in their careers who are passionate about extending the benefits of AI to communities across America. We’ll teach 1,000 fellows how to use Claude well, match them with nonprofits across America, and pay them to spend a year—full-time, in-person—helping host organizations to advance their missions. Our goals are twofold: that host organizations are equipped with valuable...

You don’t give commands. Every few ticks, the backend packages an agent's vitals, episodic memories, and grid environment, and routes it to OpenRouter (running the openai/gpt-oss-120b:free model). The LLM runs an OODA loop based on Maslow's hierarchy of needs and chooses a physical action from a structured JSON schema. They have to plant wheat, wait for it to mature, and eat it before their health hits zero. They reproduce, trade, build structures, and eventually die of old age. What actua...

Article URL: https://clawdcursor.com Comments URL: https://news.ycombinator.com/item?id=48517079 Points: 4 # Comments: 0

Article URL: https://anyframe.dev/blog/we-hired-an-intern-named-gilfoyle Comments URL: https://news.ycombinator.com/item?id=48517164 Points: 4 # Comments: 0

Article URL: https://tim.blog/2026/06/12/has-ai-already-killed-nonfiction/ Comments URL: https://news.ycombinator.com/item?id=48517169 Points: 3 # Comments: 0

Article URL: https://tselai.com/ai-elegant-software-monopoly Comments URL: https://news.ycombinator.com/item?id=48517272 Points: 3 # Comments: 0

Article URL: https://github.com/openwong2kim/wmux Comments URL: https://news.ycombinator.com/item?id=48517302 Points: 3 # Comments: 0

Article URL: https://techcrunch.com/2026/06/12/chinese-cybercrime-operation-that-used-ai-to-scam-hundreds-of-thousands-of-victims-sued-by-google/ Comments URL: https://news.ycombinator.com/item?id=48517417 Points: 2 # Comments: 0

Article URL: https://www.theregister.com/offbeat/2026/06/13/world-cup-ai-predictor-now-lets-users-ask-daft-what-ifs/5254853 Comments URL: https://news.ycombinator.com/item?id=48517421 Points: 2 # Comments: 0

Article URL: https://github.com/gambletan/cortex Comments URL: https://news.ycombinator.com/item?id=48517429 Points: 3 # Comments: 0

Article URL: https://stemteachingtools.org/brief/109 Comments URL: https://news.ycombinator.com/item?id=48517450 Points: 2 # Comments: 0

Article URL: https://www.tomshardware.com/tech-industry/cyber-security/hades-malware-campaign-now-tricks-ai-bots-by-injecting-text-about-biological-and-nuclear-weapons-failsafe-mechanisms-triggered-by-prompts-for-weapon-creation-stop-scans-before-payload-is-seen Comments URL: https://news.ycombinator.com/item?id=48517491 Points: 3 # Comments: 0

Article URL: https://fullscale.io/blog/developer-shortage/ Comments URL: https://news.ycombinator.com/item?id=48517806 Points: 2 # Comments: 0