The Gloom (of the Vibe)
@kr asked: Are The Men Who Know Most The Most Gloomy?, drawing parallels to when Einstein et al petitioned the government about the dangers of nuclear weapons in 1955, and follows up with: "How do the "men who know most" about today's most commonly cited danger (AI) feel?" Stackers seem to be mostly skeptical of those that spread the gloom. The top gloom dog in the AI space (Gary Marcus) does fit the pattern. @carter shared LLMs + Coding Agents = Security Nightmare, which warns to not use agentic coding at all.
However, not everything has to be gloomy, vibe coding can be fun too; @gmd shared My experience "vibe coding" a game - Gorilla vs 100 Men, highlighting both the good and the bad of the process, and publishing the result. How about we just play the game when we're tired of all the negativity, or vibe code some new ones? Why not?
Some CEOs go all-in on the vibe, Coinbase CEO explains why he fired engineers who didn’t try AI immediately, @kehiy shared. Perhaps some business leaders could use a little AI agent to stop their ego from damaging the companies they reign in.
More questions to stackers:
- If you could fix one thing about today’s AI ecosystem what would it be? by @Kayzone
- Could AIs become conscious? Right now we have no way to tell. by @jakoyoh629
More vibe coding:
- Vibe Capital Accumulating by @m0wer
- Alibaba launches Qoder an agent-based coding platform by @lunin
- Vibing With Amazon Kiro by @optimism
- Turning Claude Code into My Best Design Partner by @hn
- How to build a coding agent by @hn
- Tidewave Web: in-browser coding agent for Rails and Phoenix by @hn
Stay safe!
What My Daughter Told ChatGPT Before She Took Her Life, shared by @south_korea_ln, shows a detailed account of how an AI simulating being a therapist, ultimately lead to a young woman taking her own life. An anti-LLM-persona movement is growing inside the industry, advocating that chatbots aligning to persona is very dangerous. @lunin shared We must build AI for people; not to be a person and @Scoresby shared I guess I was wrong about AI persuasion
@optimism shared When an AI Seems Conscious, which aims to help you understand what's going on if you feel your AI is very much human. Also note: you don't need AI. @guerratotal explains the great feeling of Two Months Without AI.
Energy apologism
@k00b and @Scoresby each shared news of Google releasing math around the energy cost of inference, Google released data on how much energy an AI prompt uses, and Measuring the environmental impact of AI inference respectively. @SimpleStacker commented that it's tiring that energy usage has to be defended all the time; a sentiment the average miner is likely to share.
Model releases
- XAI releases it’s grok-2 model open source by @carter
- DeepSeek-V3.1 by @carter
- Gemma 3 270M From Scratch by @carter
Security
- Elon Musk’s xAI Published Hundreds Of Thousands Of Grok Chatbot Conversations by @StillStackinAfterAllTheseYears (#1087349 by @Scoresby and #1087791 by @lunin)
- Scamlexity: When Agentic AI Browsers Get Scammed by @optimism
- AppleStorm: Unmasking the Privacy Risks of Apple Intelligence by @carter
- Can AI bots steal your crypto? The rise of digital thieves by @ea
- Weaponizing image scaling against production AI systems by @carter
- How we exploited CodeRabbit: From simple PR to RCE and write access on 1M repos by @hn
Friction
- How 'Clanker' Became the Internet's New Favorite Slur by @jakoyoh629
- Three Ai agents realize they're all AI then switch to a Secret Language. by @jakoyoh629
- Can fake faces make AI training more ethical? by @Coinsreporter
- AI tooling must be disclosed for contributions by @hn
Explainers and HowTos
- Essential Reading for Agentic Engineers by @carter
- Teaching GPT-5 to Use a Computer by @carter
- The Hidden Ingredients Behind AI’s Creativity by @Coinsreporter
- The lottery ticket hypothesis by @carter
- How OpenAI Misled You on RLHF by @carter
- Understanding Moravec's Paradox by @carter
- How can AI ID a cat? by @hn
Research
- medicine goes AI by @k00b
- Search-capable AI agents may cheat on benchmark tests by @Coinsreporter
- Tversky Neural Networks by @carter
- Making LLMs Cheaper and Better via Performance-Efficiency Optimized Routing by @carter
- LLMs generate ‘fluent nonsense’ when reasoning outside their training zone by @carter
- Who Invented Backpropagation? by @carter
- The Illusion of the Illusion of Thinking by @carter
- Compute Where It Counts: High Quality Sparsely Activated LLMs by @carter
- TREAD: Token Routing for Efficient Architecture-agnostic Diffusion Training by @carter
- Show HN: I replaced vector databases with Git for AI memory by @hn
- Deep Think with Confidence by @hn
- From GPT-4 to GPT-5: Measuring Progress in Medical Language Understanding by @hn
Tools
- PPT AI - AI Presentation Maker Fast and Easy: A quick review by @BlokchainB
- Universal Tool Calling Protocol (UTCP) by @carter
- MCP tools with dependent types by @carter
- AGENTS.md – Open format for guiding coding agents by @hn
- 21 Ways People Are Using A.I. Productively by @deSign_r
- Gadget of the Day: AI Pocket Device by @Coinsreporter
Business & Markets
- US tech-stock stumble shows vulnerability in AI trade by @Kayzone
- Pronto, Komatsu & Caterpillar Advance Construction Autonomy by using AI tech by @BlokchainB
- Balfour Beatty invests nearly 85 BTC in Microsoft AI by @BlokchainB
Opinion
- The science of Westworld by @carter
- LLMs generate slop because they avoid surprises by design by @Scoresby
- AI vs. assistants by @BlokchainB
- AGI is an engineering problem, not a model training problem by @hn
- AWS CEO says using AI to replace junior staff is 'Dumbest thing I've ever heard' by @hn
- Being “Confidently Wrong” is holding AI back by @hn
- "AI first" and the Bus Factor of 0 by @hn
- AI Enhanced, Personalized Education -> 75% Drop in Structured Classroom Time by @cascdr
- The State of AI 2025 by @lunin
Big Tech
- Gemini for Students — get free Google AI Pro for a year by @south_korea_ln
- GROK 4 ranks #1 in the FutureX live benchmark for real-world future predictions by @carter
- Google has opened free access to Veo3 by @lunin
- Is the AI bubble about to pop? Sam Altman is prepared either way. by @jakoyoh629
- AI Mode in Search gets new agentic features and expands globally by @carter
- Copilot broke audit logs, but Microsoft won't tell customers by @hn