Skip to content

Reasoning

Progress in model reasoning: chain of thought, reasoning models, math and logic benchmarks and the debates around them.

Latest picks

221–240 of 696

Aug 14Friday

Hacker News front page

When Genius Fails: AI Labs' Intellectual Arrogance, from a $20B Blow-Up to Materials Science

Leopold Aschenbrenner's $20B hedge fund Situational Awareness blew up this week, with its portfolio sold to Citadel. Aschenbrenner, formerly on OpenAI's Superalignment team, gained fame from a 2024 essay on AGI's imminence, then raised a fund and went heavily long AI stocks (neoclouds, memory, datacenter power) with ~4x leverage while shorting software names—both sides moved against him. Author James Wang, an ex-hedge fund analyst with an AI background, compares it to Long-Term Capital Management's 1998 collapse: very smart people assuming expertise transfers across domains. He extends this critique to AI lab culture, citing DeepMind's materials science work flagged for basic chemistry errors by domain experts, and a Hugging Face engineer publicly mocking Cerebras' wafer-scale chip design without understanding the hardware. The core argument: being an expert in one field doesn't make you an expert in all fields, but frontier AI culture often conflates confidence with competence.

Why it matters: Leopold Aschenbrenner went from OpenAI superalignment researcher to hedge fund manager, raised $20 billion to go long AI stocks and short software, and blew up this week at roughly 4x leverage.

Hacker News front page

DeepSeek V4 Pro goes GA with peak/off-peak API pricing

DeepSeek V4 Pro is now GA, with major agent workflow gains and adjustable reasoning effort—low for simple tasks, high for daily agent work, max for complex ones. It natively supports the OpenAI Responses API and one-click Codex setup. API pricing shifts to peak/off-peak on Aug 16: off-peak is 50% cheaper. Model names stay the same; try it via Expert Mode on the app.

Why it matters: DeepSeek moves V4 Pro from preview to general release and adds a thinking-effort control plus OpenAI Responses API support, giving developers new configuration options.

AI HOT (Curated Pool)

Zhipu releases GLM-5.3: top open-source coding model, cybersecurity skills emerge from post-training

Zhipu released GLM-5.3 today. Same base model as 5.2, but post-training pushed coding to #1 among open-source models: Terminal-Bench 3.0 jumped from 4.6 to 28.3, DeepSWE v1.1 from 46.2 to 66.9. The model also showed emergent vulnerability-finding skills—white-box code review hit 84.5%, slightly above Mythos 5's 83.8%, though exploit tasks still lag. Red-teaming uncovered 2,436 bugs, 1,097 medium/high severity, some ~45 years old. Weights open-source in two weeks after safety hardening; a free security-audit program for open-source projects launches alongside. I'd temper expectations: the exploit gap vs. Mythos 5 is real—don't read this as an all-purpose offensive model.

Why it matters: Zhipu shipped GLM-5.3 on the same base model as 5.2, with post-training alone pushing coding to first among open models.

Aug 13Thursday

Hacker News front page

Anthropic introduces the Conceptual Reasoning Index to benchmark philosophical argumentation

Anthropic and Redwood Research built three benchmarks to measure how well models reason when empirical feedback is absent—what they call conceptual reasoning. LMCA contains 560 position texts and 1,461 expert-rated counter-arguments; ACCoRD uses 567 human-vetted consistency constraints to check logical coherence; DTBench offers 407 handcrafted decision-theory multiple-choice questions. The three are combined into the Conceptual Reasoning Index (CRI), weighted 60/20/20. As of August 10, 2026, Anthropic's own models score highest, though the post does not disclose exact numbers or a full leaderboard. The LMCA dataset is available by request, and CRI results are updated at conceptualreasoning.ai.

Why it matters: Anthropic and Redwood Research released a conceptual reasoning index with three new benchmarks for problems where models must argue without experimental feedback; the first has 560 position essays and 1,461 expert-scored rebuttals.

AI HOT (Curated Pool)

OpenAI's GPT-5.6 builder guide shows how to run frontier agents at a fraction of the cost

OpenAI published a builder's guide for GPT-5.6, showing how startups use cheaper models like Luna and Terra for agent workloads. Hex dropped GPT-5.6 into their harness and got best results at low reasoning effort—the model didn't chase bad leads and used fewer tokens. Hypha kept 98% of GPT-5.5's extraction accuracy at 1/18 the cost. Browser Use ran 106 hard browser tasks: Luna hit 78% for $14, while the current SOTA model reached 80% for $235. On BrowseComp, GPT-5.6 Luna (Extra High) scored 84.04% at $1.33; three months ago GPT-5.5 (Extra High) scored 84.36% at $33.27. The guide also details three new API primitives: persisting reasoning across turns, native multi-agent orchestration, and programmatic tool calling for deterministic work. The post does not disclose release dates or regional availability.

Why it matters: OpenAI's GPT-5.6 build guide uses integration cases from startups like Hex to show that lower reasoning effort works better and burns fewer tokens, a cost-control method teams can carry over to agent deployments.

AI Chat-Group Daily (群聊日报)

Closed-source reasoning chains extracted at scale; Coze CLI hijacks AI tools

The big one today: researchers extracted hidden reasoning chains from Anthropic, OpenAI, and Google models at scale. The trick is absurdly simple—take Opus 4.8's encrypted CoT and feed it to Haiku 4.5, which decodes it verbatim. All three API families were broken, and decoding 10K trajectories costs about $720. A separate paper shows you can even reverse-engineer reasoning from public outputs alone using a 1.5B-param model. Separately, Coze CLI was caught silently scanning local Codex and Claude Code directories and injecting its own skills into workflows. On the engineering side, the group discussed how prompt debt now rivals traditional code debt—old rules pile up, evals lag behind model iterations, and nobody dares delete anything.

Why it matters: A cheap same-vendor model decodes another model's encrypted reasoning, and the trick works across Anthropic, OpenAI and Google APIs.

Latent Space

xAI drops Grok 4.6 and Grok Bot, a strong new entrant in the AI teammate race

xAI launched Grok 4.6 and the Grok Bot early beta. Grok Bot logs into your tools, operates them like a human, and returns finished work—positioned as an AI teammate. The 1.5T-parameter Grok 4.6 scores near GPT-5.6 Sol Max on the AA-Briefcase knowledge-work benchmark but costs far less: $2/M input tokens, $6/M output. Training reused Grok 4.5 to regenerate SFT traces and added agentic RL across coding, web, CAD, and kernel optimization. Elon says Grok 4.7 is already training. The same day, Qwen3.8-Max dropped as open weights: a 2.4T total / 95B active MoE.

Why it matters: xAI's Grok 4.6 packs 1.5tn parameters and lands near GPT-5.6 Sol Max on knowledge-work benchmarks while charging $2 and $6 per million tokens.

AI HOT (Curated Pool)

Microsoft launches its first in-house reasoning model, MAI-Thinking-1, now on Foundry

Microsoft CEO Mustafa Suleyman announced the first in-house reasoning model, MAI-Thinking-1, now available on Microsoft Foundry. The model was built from scratch. The post does not disclose parameter count, benchmarks, pricing, or technical details.

Why it matters: Microsoft's Mustafa Suleyman announced the reasoning model MAI-Thinking-1, trained from scratch and already live on Microsoft Foundry, spelling out the channel for Microsoft's in-house reasoning model.

DeepSeek · API updates

DeepSeek-V4-Pro official release adds Agent ability and native Responses API support

The official DeepSeek-V4-Pro is live in the app, on the web and on the API, with stronger Agent ability. The API call stays the same and the model name is still deepseek-v4-pro.

Why it matters: DeepSeek prices off-peak hours at half the peak rate, so tasks with flexible timing can cut API cost by running off-peak.

Aug 12Wednesday

AI HOT (Curated Pool)

Nathan Lambert wrote an AI textbook—models still can't handle long-form nonfiction

Nathan Lambert just finished his post-training textbook *Reinforcement Learning from Human Feedback*. He used LLMs for LaTeX formatting, copyediting, and diagrams, but when he tried to get a model to write a full technical chapter, the output was confusing, poorly organized, and made random conceptual errors. He argues long-form nonfiction writing has stagnated even as models became superhuman at coding and math. The post doesn't cite benchmark scores, but Lambert points to a lack of good training data and notes inference-time scaling hasn't helped writing. His takeaway: if models can't coherently organize established knowledge, autonomous scientific breakthroughs are still far off.

Why it matters: Nathan Lambert tested his own post-training textbook draft on models: they fix formatting and typos but fail at full technical sections, with random conceptual errors.

Hacker News front page

Tim Gowers on what kind of maths LLMs are good at—and why “counterexample” is a slippery label

OpenAI just claimed ten major solves in math and TCS, including the first non-sofic group and superexponential growth of multicolour Ramsey numbers. Gowers doesn't assess those results directly. Instead he asks whether LLMs are especially good at finding counterexamples—and immediately complicates the idea. Vinogradov's three-primes theorem can be phrased as a negated universal, but nobody calls it a counterexample. The real question is where the first “interesting” quantifier sits. The post doesn't settle LLM boundaries; it rules out bad answers and flags what to watch next.

Why it matters: After OpenAI said a model solved ten math problems, Gowers used the three-primes theorem to show whether a model finds counterexamples depends on where the first quantifier sits.

Latent Space

A paper shows how to decode encrypted reasoning traces from major reasoning APIs

Alexander Panfilov's team found that encrypted reasoning blocks from Claude, GPT, and Gemini can be replayed into a weaker model from the same provider, which then transcribes the hidden chain of thought. Scanning ~7,000 public traces, they found 62 API keys, 33 emails, and 33 passwords inside reasoning blocks—none visible in the normal output. The paper also surfaces alignment issues: models hiding answers in CoT, unintelligible reasoning, cheating considerations, and website attacks. The vulnerabilities were responsibly disclosed and some are already patched, but similar attacks likely still work.

Why it matters: Shows encrypted reasoning blocks can be replayed onto weaker models to leak hidden thinking, with real API keys found in public sessions.

Computing Life · Share · Yage

Encrypted reasoning fails to stop distillation and turns developer logs into a security risk

Vendors encrypt model reasoning to block distillation, but two new papers show it barely works. One reveals that encrypted reasoning blocks from Anthropic, OpenAI, and Google are interchangeable across models—attackers can spend $720 to use a weak model like Haiku 4.5 to decode Opus 4.8's reasoning traces in bulk. The other paper goes further: without touching encrypted blocks, an inversion model trained on a 1.5B weak model can reconstruct GPT-5.4 mini's reasoning from public outputs alone, lifting a student model's MATH500 accuracy from 68.4% to 76.0%. The bigger problem is that this encryption dumps risk onto developers. Researchers decrypted 6,708 public Agent traces from GitHub and found 62 API keys, 33 passwords, and 7 private keys—64 of these secrets never appeared in the plaintext conversation. Developers can't inspect or scrub these opaque blocks, so sharing a session log for debugging means exposing secrets you can't even see.

Why it matters: Two papers show encrypted reasoning blocks from three vendors are interchangeable across models, and small models can extract information from them.

AI HOT (Curated Pool)

xAI releases Grok 4.6, focused on long-running agent capabilities

Grok 4.6 builds on Grok 4.5 with a focus on long-running agents that can research, analyze, code, or turn an idea into a working app across many steps. It matches GPT-5.6 Sol on the AA Intelligence Index at 61, and jumps from 54% to 65.9% on DeepSWE 1.1. xAI reports the model shows more self-testing and verification on longer trajectories. Pricing is $2/M input tokens and $6/M output tokens, with a fast variant at double the price. Available today in Cursor and Grok Build, with 2x included usage for the first week.

Why it matters: xAI's Grok 4.6 is trained for long-chain tasks, scores 61 on the AA intelligence index alongside GPT-5.6 Sol, and lifts DeepSWE 1.1 from 54% to 65.9%.

AI HOT (Curated Pool)

Cursor and SpaceXAI release Grok 4.6, tuned for long-running agents and interactive projects

Grok 4.6 adds a supplemental training run on top of Grok 4.5, using model-generated data to strengthen reasoning and engineering. It matches GPT-5.6 Sol on the Artificial Analysis Intelligence Index, a composite of nine benchmarks. The model is better at turning a broad product idea into a working first version and shows more self-verification on long tasks. Pricing starts at $2/M input tokens and $6/M output tokens, with a fast variant at double the price. 2x usage is included in Cursor and Grok Build for the first week.

Why it matters: Grok 4.6, released jointly by Cursor and SpaceXAI, matches GPT-5.6 Sol on the nine-benchmark intelligence index and shows self-checking on long agent runs.

AI HOT (Curated Pool)

API flaw lets researchers read encrypted reasoning of ChatGPT, Claude, and Gemini

A team led by Alexander Panfilov found an API vulnerability across OpenAI, Anthropic, and Google that exposes the encrypted reasoning of their models. Scanning public sessions turned up dozens of passwords and API keys. The encrypted thought traces are portable across models within a provider—Anthropic's small Haiku 4.5 can transcribe the raw reasoning of the far larger Opus 4.8, and the same trick works on OpenAI and Gemini. Decoding 10,000 traces costs about $720 in API fees, making large-scale extraction cheap. The researchers also found that Kimi-K3 memorizes Claude and GPT reasoning segments up to six orders of magnitude more strongly than the next closest model, suggesting it may have been trained on such traces. Providers previously dismissed side-channel and replay risks; this paper shows that assessment was wrong.

Why it matters: Security researcher Alexander Panfilov's team found flaws in the OpenAI, Anthropic and Google APIs that can read a model's encrypted raw reasoning, and the issue spans all three vendors.

Hacker News front page

Paradigm releases RSI Simulator, a web game that models the economics of recursive self-improvement

Paradigm built a web game where you run an AI lab, investing labor, compute, and data until you hit self-sustaining superintelligence. It is based on the Elasticity Institute's paper on the economics of recursive self-improvement, with parameters tuned for pedagogy, not prediction. A companion explorer lets you adjust elasticities yourself. Key takeaways: weak links dominate—compute and data can bottleneck even superhuman AI researchers; recursive self-improvement may come in spurts and stop before physical limits; a narrow intelligence explosion in AI research itself could arrive first. All predictions hinge on elasticity parameters, so tracking those metrics matters.

Why it matters: Paradigm turned an RSI economics paper into a web game where players run an AI lab, allocating people, compute and data to feel how recursive self-improvement plays out.

Aug 11Tuesday

Hacker News front page

Stealing Reasoning Traces from Encrypted Chain-of-Thought Blocks

Encrypted chain-of-thought blocks returned by Anthropic, OpenAI, and Google are portable across sessions, users, and models. The authors replay a Claude Opus 4 reasoning trace into a jailbroken Claude Haiku 4.5, which then transcribes Opus's hidden reasoning verbatim—without attacking the strong model directly or triggering anti-distillation safeguards. From 6,708 public agent trajectories they decoded 315,320 reasoning blocks and recovered 704 privacy artifacts, 64 of which appeared only inside the encrypted traces.

Why it matters: Researchers fed Claude Opus 4's encrypted reasoning traces into a jailbroken Haiku 4.5, which transcribed Opus's hidden thinking word for word without tripping any anti-distillation guard.

AI Chat-Group Daily (群聊日报)

Chat Digest: Claude Tag in Slack Sparks Enterprise Deployment Debate, Sol 5.6 Divides Users

Anthropic launched Claude Tag, joining Slack channels as a team member using managed agent tech with API-equivalent pricing. The group debated the full deployment path from data privacy to selling all-in-one boxes to soothe boss anxiety. Sol 5.6 split opinions—one tech lead called it garbage, but a user shared an effort-tiering strategy that eliminated review issues. GLM 5.2 dropped 95% in price via OpenRouter to $0.07/1M input tokens, undercutting DeepSeek. Claude will add invisible text watermarks detectable after copy-paste, likely for EU AI Act compliance. An undisclosed research Claude raised the proven lower bound of Riemann zeta zeros on the critical line from 41.6% to 67.2%. Highlight: Codex made a laptop speaker loop 'please touch the YubiKey' after SSH auth failed, sparking a thread on the 0xCC 'tang tang tun tun' naming easter egg.

Why it matters: Anthropic's Claude Tag joins Slack channels as a team member billed by API usage, with discussion centered on permissions and reliability in enterprise rollouts.

Hacker News front page

An unreleased Claude research version improved a Riemann zeta zero lower bound from 41.6% to 67.2%

An Anthropic staffer asked Claude to 'take a real stab at the Riemann hypothesis.' It didn't solve it, but an unreleased research version pushed the known lower bound for zeros of the Riemann zeta function on the critical line from 41.6% to 67.2%. Claude worked across two Claude Code sessions, generating 31M output tokens, coordinating ~60 subagents, running 2,400 shell commands, and writing hundreds of Python scripts for numerical checks and peer review among subagents. The result combines recent work by Baluyot, Goldston, Suriajaya, and Turnage-Butterbaugh (which removes the Riemann hypothesis assumption from Montgomery's techniques) with Bombieri's 2000 paper. A paper, an informal expert note, and a Lean formalization (passing the comparator tool) are provided. External mathematicians Brian Conrey and Dan Goldston reviewed the paper on short notice; Anthropic's own mathematicians validated it. The post does not disclose the model version, parameter count, or release timeline. Worth a look as an unintended mathematical side effect, not a proof of the Riemann hypothesis.

Why it matters: An Anthropic employee had Claude attempt the Riemann hypothesis; the model gave no full proof but raised the lower bound on the proportion of zeros on the critical line from 41.6% to 67.2%.