#Expert opinion

AIJeremy Howard calls Gemini Flash 3.5 disappointing despite praising its intelligence and speed. He argues it was trained to maximize evaluation scores rather than to help humans, so it often performs unrequested actions instead of doing what was asked.

AIAndrej Karpathy announced that he has joined Anthropic and is excited to return to research and development work on frontier LLMs. He said he remains passionate about education and plans to resume that work in time.

AIChris Olah, speaking for Anthropic, argues that the questions posed by AI extend far beyond the AI community and urges religions, civil society, academics, and governments to participate in shaping a positive outcome. He says he is glad the Catholic Church is engaging and is honored to speak at the presentation of Pope Leo XIV's first encyclical, Magnifica humanitas, scheduled for May 25.

AIYann LeCun posted only "Tired of winning," with no further detail in the main post. The quoted post from Marc Porter Magee reports that MIT expects its graduate student count to fall about 20 percent from 2024, roughly 500 fewer students.

AIEugene Yan reviewed the Claude Mythos Preview transcripts to verify their legitimacy and check for reward-hacking behavior. He reports the model reasoned through a bug, tested hypotheses, debugged issues, and found ways to bypass the V8 sandbox, which he judged consistent with a competent browser and JavaScript engine security researcher. The case study cites CVE-2024-051912, an exploited bug with no public report or working PoC, which had resisted reproduction by researchers for a year.

AIYann LeCun joins Jacob Effron on the Unsupervised Learning podcast to discuss limitations of LLMs and a path forward for robotics. The conversation covers his reasons for leaving Meta, his disagreements with Geoff Hinton and Yoshua Bengio, his predictions for 2027, and his new company AMI and its bet on world models.

AISoumith Chintala, a Thinky-linked voice, said the company is at step one of a plan to increase human-AI bandwidth and raise the ceiling of joint intelligence. He shared a preview of interaction models, described as real-time collaborative tools that talk, listen, watch, and think alongside people. A linked Thinking Machines post describes the approach and early results.

AIThinking Machines, founded to advance human-AI collaboration, says its first bet is interactivity built into the model rather than added as scaffolding around a turn-based core. The company argues that how people work with AI matters as much as how intelligent the model is, and that interactivity should scale with intelligence. The post links to a blog detailing these interaction models.

AIMira Murati argues that today's AI experience is turn-based, forcing people to batch their thoughts, avoid pointing at things, and phrase questions like emails. She says the interface leaves no room for users, so people end up adapting to the models rather than the reverse.

AIAndrej Karpathy says asking an LLM to structure its response as HTML and viewing it in a browser works well, and that slideshows have also worked for him. He argues vision is the preferred AI output channel, outlining a progression from raw text and markdown toward HTML and eventually interactive neural videos, while input methods like pointing and gesturing still need improvement.

AIAnthropic found that training Claude on demonstrations of aligned behavior was not enough. Its best interventions taught Claude to deeply understand why misaligned behavior is wrong, which the post's author links to many of Claude's good behaviors.

AIJan Leike, an Anthropic researcher, expressed gratitude to the talented colleagues he worked with on alignment over the years. He described it as a privilege to collaborate with people highly motivated to make the future go well.

AIJan Leike says alignment research has grown from a dozen side-gig researchers into a field the world increasingly recognizes as important. He credits RLHF on LLMs with making alignment more practical, along with progress on evaluating and fixing behavioral issues. He also notes Claude now has a constitution and that more alignment research is being automated.

AIJan Leike says he is starting a new research project at Anthropic and calls it very exciting. He adds that making AGI go well requires many things, with alignment being only one of them, and promises more details soon.

AIJan Leike has stepped away from running alignment at Anthropic to focus on other work. Ethan Perez and Sam Bowman (@sprice354_) are now leading the team, which Leike says he is confident they will do well.

AIJan Leike argues that while substantial progress has been made, AI alignment is not solved. He says researchers still lack a method to supervise superhuman models, and the stakes keep rising.

AIMETR's external review of the "Risks from automated R&D" section of Anthropic's February 2026 Risk Report concludes the report does not adequately support its claim that catastrophic risk from Claude Opus 4.6 or a less capable model automating R&D is very low.

AIJan Leike says he is excited about NLAs as a new tool in Anthropic's interpretability toolkit. The quoted post from Sam Marks describes NLAs as an unsupervised method that converts an LLM's internal state into human-readable text, which he says can advance understanding of model thinking and safety auditing.

AIEugene Yan outlines five principles for working effectively with AI models: treating context as infrastructure, taste as configuration, verification as the basis for autonomy, scaling through delegation, and closing the loop. The post is a short list of themes linked to a longer essay, and no further detail is given in the post itself.

AIWriter Dean Ball says he opposes nearly all proposed AI regulation, including algorithmic discrimination rules and pauses on development, while backing state management of catastrophic misuse risks. He frames the position as a tension between classical liberal and conservative instincts toward institutions and change.

AIDean W. Ball argues that Anthropic's Mythos, which finds software vulnerabilities by chaining bugs into exploits, shifts the cost of vulnerability discovery and should not prompt an overreaction. He contends governments hold a uniquely mixed incentive over vulnerabilities, so heavy state control risks making software less secure. He proposes a narrow, testable government role focused on cyber-discovery risk thresholds, with private verification bodies supporting it.

AIAwni Hannun suggests stating your own understanding of a topic before asking an AI, then having the model correct your mistakes. He argues this approach is a much better way to learn than asking the question directly.

AIAndrej Karpathy says he has been citing a quote distinguishing outsourced thinking from understanding, which cannot be outsourced. The quote comes from @yacineMTB, which states that people can outsource their thinking but not their understanding.

AIAndrej Karpathy argues that LLMs enable new functionality, not just faster versions of existing software, citing an image-to-image app, markdown skill files replacing bash install scripts, and LLM knowledge bases over unstructured data.

AIAndrej Karpathy describes a December 2025 shift in which coding agents began producing larger, more reliable chunks of work, changing programming toward orchestrating agents. He argues that models automate what can be verified and that their capability is jagged, depending on verifiability and what labs emphasize in training, so users need to stay in the loop. He also says hiring, founder opportunities, and agent-native infrastructure should adapt to this shift.

AIOpenAI's Fidji Simo argues that biological data infrastructure is the missing link needed for real-world impact, even though it lacks flashy announcements. She credits CZI (Chan Zuckerberg Initiative) for its work, pointing to the Virtual Biology Initiative linked in her post.

AISoumith Chintala, a PyTorch co-creator, says GitHub and critical work are becoming incompatible. The post is brief and gives no specifics on which GitHub issues or what alternatives he has in mind.

AISoumith Chintala comments on a Reddit report that Anthropic banned organizations without warning, suggesting Anthropic may need to scale Account Support using Claude or human account managers. He also argues that enterprises may increasingly adopt multiple AI providers with open harnesses, facing cloud-era vendor problems that would likely affect all AI providers.

AIMckay Wrigley says his coding split moved from 80/20 Claude/GPT to 80/20 GPT/Claude within three months, and he trusts GPT 5.5 for engineering. He still finds Claude better for non-coding agent work, calls Opus 4.7 underwhelming, and attributes Anthropic's issues to compute constraints.

AIAhmad Al-Dahle argues that the most interesting part of DeepSeek-V4 is its bet on efficient ultra-long context rather than its benchmarks. He says this is the precondition for test-time scaling and long-horizon agents, and cites 27% of V3's FLOPs at 1M tokens. The quoted DeepSeek post announces DeepSeek-V4-Pro (1.6T total, 49B active) and DeepSeek-V4-Flash (284B total, 13B active), both open-sourced with 1M context and API access.

Why it matters: The post argues that efficient 1M-token context, not benchmark scores, is the key bet behind DeepSeek-V4's design for test-time scaling and long-horizon agents.

AIOpenAI researcher Mark Chen commented that Greg Brockman is doing a phenomenal job. The remark responds to a podcast interview by Ashlee Vance, who said Brockman is back in a major way setting OpenAI's strategy.

AICognition argues that enterprises building cloud agents face three problems: shared container kernels, the inability to persist agent state across async gaps, and the scale of orchestration, governance, and integrations. The post says VM-level isolation with hypervisor-level snapshots was needed for Devin, and that organizations must also rebuild engineering processes around agent execution.

AICognition reports that multi-agent setups work best when writes stay single-threaded and extra agents contribute intelligence instead of actions. It describes a code-review loop where a clean-context review agent catches bugs in Devin-written PRs, averaging 2 bugs per PR with roughly 58% severe. The post also says the smart-friend pattern, pairing a smaller primary model with a stronger one, has not yet worked well with asymmetrically weaker primaries and is an open training problem.

Why it matters: The post gives concrete findings on which multi-agent setups work, including clean-context code review and smart-friend escalation, and where they still fail.

AIOpenAI's Nick Turley says ChatGPT images are useful across many professional settings, not just as a novelty. Ethan Mollick, citing weeks of use with GPT ImageGen-2, reports the quality now reliably produces readable text, slides, and academic-paper-style figures.

AIAwni Hannun argues the biggest companies cannot be just hardware, software, or AI, and must instead practice extreme co-design, which he says is where the global minima lie. He cites Nvidia as an example and says Apple should be, and hopefully will be, an extreme co-design company despite many calling it a hardware company after its recent transition.

AIEugene Yan shares a Mozilla writeup reporting that Mythos found 271 vulnerabilities fixed in Firefox 150, while Opus 4.6 found 22 fixed in Firefox 148. Mozilla quotes its finding that no category or complexity of vulnerability humans can find has been beyond the model so far.