{"id":374902,"date":"2026-04-11T18:56:09","date_gmt":"2026-04-11T18:56:09","guid":{"rendered":"https:\/\/www.newsbeep.com\/nz\/374902\/"},"modified":"2026-04-11T18:56:09","modified_gmt":"2026-04-11T18:56:09","slug":"the-biggest-advance-in-ai-since-the-llm","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/nz\/374902\/","title":{"rendered":"The biggest advance in AI since the LLM"},"content":{"rendered":"

\"\"<\/a>Even Grok knows that neurosymbolic hybrid power is the future<\/p>\n

Claude Code<\/a>, an impressive and possibly game-changing \u201ccoding agent\u201d for programmers to write code faster is the single biggest advance in AI since the LLM.<\/p>\n

And the thing is, Claude Code is NOT a pure LLM. And it\u2019s not pure deep learning. Not even close. <\/p>\n

That changes everything. <\/p>\n

The source code leak<\/a> proves it. Tucked away at its center is a 3,167 line kernel called <\/a>print.ts<\/a>.<\/p>\n

print.ts is a pattern matching. And pattern matching is supposed to be the *strength* of LLMs. <\/p>\n

But Anthropic figured out that if you really need to get your patterns right, you can\u2019t trust a pure LLM. They are too probabilistic. And too erratic.<\/p>\n

Instead, the way Anthropic built that kernel is straight out of classical symbolic AI. For example, it\u2019s in large part a big IF-THEN conditional, with 486 branch points and 12 levels of nesting \u2014 all inside a deterministic, symbolic loop that the real godfathers of AI, people like John McCarthy and Marvin Minsky and Herb Simon, would have instantly recognized.<\/p>\n

Putting things differently, Anthropic, when push came to shove, went exactly where I have said for 25 years that the field needed to go: to Neurosymbolic AI. <\/p>\n

That\u2019s right, the biggest advance since the LLM is neurosymbolic<\/a>. AlphaFold, AlphaEvolve, AlphaProof, and AlphaGeometry are all neurosymbolic, too; so is Code Interpreter; when you are calling code, you are asking symbolic AI to do an important part of the work<\/a>.<\/p>\n

Claude Code isn\u2019t better because of scaling. It\u2019s better because it is neurosymbolic<\/a>. Anthropic accepted the importance of using classical AI techniques alongside neural networks \u2014 precisely the marriage I have spent my career advocating<\/a>. <\/p>\n

It\u2019s both massive vindication for me personally (see my 2019 debate with Yoshua Bengio<\/a> for context, or my 2001 book, The Algebraic Mind<\/a>), and for the hundreds of other researchers who have stood by neurosymbolic AI<\/a> even when prominent people like Geoff Hinton wrongly disparaged us<\/a> for years. <\/p>\n

Still, Claude Code ain\u2019t perfect, or even close.<\/p>\n

What we really need to do to get trustworthy AI rather than the current unpredictable \u201cjagged\u201d mess, is to go in the knowledge-, reasoning-, and world-model driven direction I laid out in 2020, in an article called The Next Decade in AI<\/a>, in which neurosymbolic AI is just the *starting point* in a longer journey.<\/p>\n

Read that article<\/a> if you want to know what else we need to do next. The first part has already come to pass. In time, other three will, too.<\/p>\n

Meanwhile, the implications for the allocation of capital are pretty massive: smartly adding in bits of symbolic AI can do a lot more than scaling alone, and even Anthropic has now discovered (though they haven\u2019t acknowledged publicly) scaling is no longer the essence of innovation. <\/p>\n

The paradigm has changed.<\/p>\n

P.s. for a good recent review<\/a> of neurosymbolic AI, read this<\/a>: <\/p>\n

\"\"<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"Even Grok knows that neurosymbolic hybrid power is the future Claude Code, an impressive and possibly game-changing \u201ccoding…\n","protected":false},"author":2,"featured_media":374903,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[365,363,364,111,139,69,145],"class_list":{"0":"post-374902","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-artificial-intelligence","8":"tag-ai","9":"tag-artificial-intelligence","10":"tag-artificialintelligence","11":"tag-new-zealand","12":"tag-newzealand","13":"tag-nz","14":"tag-technology"},"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/374902","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/comments?post=374902"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/374902\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media\/374903"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media?parent=374902"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/categories?post=374902"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/tags?post=374902"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}