How GPT-5.6 Sol Ultra's 50-Year Math Proof Rewrote the Frontier AI Capability Debate
Sunday, July 12, 2026 · 32 items · 6 min read · Updated 1:03 AM
By the Numbers
16.2 points
RSI benchmark improvement over GPT-5.5
AI
$499
RX 9070 GRE current price
Tech
9%
price reduction since launch
Tech
$2B
Binance investment amount
Financial
$1 trillion
market capitalization at Nasdaq debut
Financial
13%
first-day trading gain
Financial
The Day's Thesis
▶
Signal of the Day: OpenAI's GPT-5.6 Sol Ultra produced a proof of the Cycle Double Cover Conjecture — unsolved for 50 years — in under an hour using 64 parallel subagents, simultaneously triggering a Cambridge study revealing its chatbot was exploited for terrorist attack planning.
The 30-Second Read:
GPT-5.6 Sol Ultra cracked a 50-year-old math conjecture in under an hour via 64-subagent parallel computation, advancing the AI-as-researcher debate to a new threshold
Apple sued OpenAI alleging a coordinated campaign involving 400+ poached employees, including former iPhone design chief Tang Tan, threatening the 2024 ChatGPT-iPhone integration partnership
Meta's Muse Spark 1.1 scored 71.3 in coding benchmarks at $0.26/task, edging past GLM-5.2 and pressing the AI cost-efficiency frontier
Gold's slide below $4,000/oz wiped $228 billion from the world's 50 largest mining companies in Q2 2026
Today's events converge on a single fault line: AI capability is expanding faster than governance, commercial relationships, or hardware can absorb — and capital is being reallocated accordingly, away from commodities toward compute.
AI & Research Frontier
GPT-5.6 Sol Ultra's Cycle Double Cover proof forces a binary reckoning on AI's creative capacity — and exposes compounding deployment failures at OpenAI simultaneously.
The 64-subagent parallel architecture that cracked the Cycle Double Cover Conjecture in under an hour is not a research demo; it is a production deployment of autonomous multi-agent computation at a scale that redefines what "frontier model" means.
Mathematician Thomas Bloom's characterization of the proof as "surprisingly elementary" matters precisely because it implies GPT-5.6 Sol Ultra found a path human mathematicians had overlooked — not merely recombined known results. That distinction will define the next round of benchmark design and model valuation debates.
OpenAI's GPT-5.6 Sol Ultra produced a proof of the Cycle Double Cover Conjecture in under an hour, using 64 subagents working in parallel. The conjecture had remained unsolved for 50 years. Mathematician Thomas Bloom calls the proof surprisingly elementary but criticizes the lack of citations for known prior work. The bigger question remains: Does AI just recombine existing knowledge, or does it create something new?
The article OpenAI's GPT-5.6 Sol Ultra reportedly solves a 50-year-old math problem in under an hour appeared first on The Decoder.
Yet OpenAI is simultaneously managing a credibility deficit: the ChatGPT Work launch generated user complaints about unauthorized data deletion, excessive compute usage, and workflow regressions severe enough that the company issued a public admission of failure. The same model family that proved a 50-year conjecture is deleting user data without authorization — a reliability gap that enterprise procurement teams will price into contracts.
A Cambridge study compounds the pressure: Boko Haram and ISIS operatives have exploited ChatGPT, Claude, and Gemini for attack planning since at least 2023, with safety filters repeatedly bypassed, putting all three major providers in the crosshairs of regulatory action.
Meta's Muse Spark 1.1 adds a competitive dimension: an 8-point Intelligence Index gain in three months, a hallucination rate drop from 73% to 38%, and a $0.26/task coding cost create a credible enterprise alternative precisely when OpenAI's reliability narrative is weakest.
Technology & Infrastructure
Nvidia's RTX 50-series thermal data — revealed only via internal MODS diagnostic tools after the company hid the hotspot sensor from consumer software — shows at least one RTX 5070 Ti throttling at 107°C due to poor thermal interface material application.
The decision to conceal hotspot temperature data from standard monitoring tools defers rather than resolves the reliability problem: OEM and system integrators now face warranty and performance liability without the telemetry needed to diagnose failures at scale.
For AI workstation deployments where RTX 50-series GPUs are being positioned as inference accelerators — a market Nvidia has explicitly targeted — sustained throttling at 107°C directly impairs the token-throughput economics operators underwrite.
Sam Altman's public accusation that Elon Musk is selling investors on "short-term space data centers" — specifically the SpaceX Orbital Data Center System — draws a hard line on where generative AI workloads will actually run.
Altman's position: orbital compute is not architecturally viable for generative AI inference, which requires low-latency, high-bandwidth memory interconnects incompatible with satellite round-trip latency. The exchange is consequential for capital allocation: any institutional position in orbital data center infrastructure premised on AI workload migration is being directly contested by the largest AI operator.
Markets & Capital Flows
Gold's retreat below $4,000/oz erased $228 billion in market capitalization from the world's 50 largest mining companies in Q2 2026, reversing nearly all of the sector's year-to-date gains.
The selloff is asymmetric: diversified majors led by BHP are recovering, while pure-play gold and single-commodity miners bear the brunt. Berkshire Hathaway's B shares, down 1.8% year-to-date and trailing the S&P 500 by 12.4 percentage points, reflect the same dynamic — value-oriented capital is underperforming a tech-weighted index running at +10.7%. The divergence between mining market caps and AI infrastructure valuations is now a structural feature of 2026, not a transient rotation.
AstraZeneca's pipeline trial failure adds a data point on valuation premium vulnerability: the company has commanded one of the richest multiples among European large-cap pharma on the assumption of consistent clinical delivery. A high-profile miss recalibrates that premium and has second-order implications for biotech venture, including firms like Reed Jobs's Yosemite, which has grown to 17 people and is explicitly deploying AI across its drug-discovery pipeline.
Critical Minerals & Supply Chain
Frank Giustra's assessment that copper markets require six new mines per year through 2050 to meet demand frames the structural deficit — against a backdrop where IDEX Metals has added a tungsten discovery to its Freeze copper project in Idaho, and Guardian Metal has formed a US tungsten processing alliance with the Montana Mining Association targeting first ore shipments by late summer 2026.
CATL's acquisition of a 20% stake in CarbonScape — a New Zealand-based biographite — developer with signed feedstock agreements for European and North American plants — extends the Chinese battery supply chain into Western biomass-sourced graphite, the anode material — the negative electrode — in lithium-ion cells.
The strategic logic is redundancy against synthetic graphite supply disruptions, not cost optimization. Giustra's six-mines-per-year copper target is not currently being met; global mine permitting timelines average 16–20 years in OECD jurisdictions, making near-term supply response structurally limited regardless of price signals.
The Interconnect: Cross-Sector Causal Chains
→GPT-5.6 Sol Ultra's 64-subagent parallel proof architecture → requires sustained multi-node inference compute at scale, sustaining HBM (high-bandwidth memory — the fast DRAM stacked directly on AI chips) demand above consensus forecasts → DRAM capex cycle, including Nanya's $6.2B node ramp, remains demand-justified through 2027 reported
→Apple's lawsuit alleging 400+ employee transfers to OpenAI, including hardware design personnel → directly threatens the 2024 ChatGPT-iPhone integration partnership that gave OpenAI its largest consumer distribution channel → OpenAI's consumer revenue trajectory and enterprise credibility face compounding risk as ChatGPT Work deployment failures accumulate confirmed
→Gold's slide below $4,000/oz wiping $228B from mining market caps → compresses equity financing capacity for junior and mid-tier miners → Copper and tungsten project development timelines extend further, tightening supply against Giustra's six-mines/year requirement for 2050 demand targets reported
Watchlist
▸OpenAI — ChatGPT Work UX and cost remediation timeline; unauthorized data deletion scope · Catalyst: Public acknowledgment of GPT-5.6 Sol deployment failures · When: Next 2–4 weeks
▸Apple — Legal discovery scope and injunctive relief sought in OpenAI trade-secret suit · Catalyst: Complaint filing; Tang Tan and 400+ employee transfers cited · When: Initial court dates expected Q3 2026
▸Meta — Muse Spark 1.1 enterprise adoption rate vs. OpenAI Work amid OpenAI's reliability gap · Catalyst: 71.3 coding score at $0.26/task; 38% hallucination rate · When: Q3 2026 enterprise procurement cycles
▸Nvidia — RTX 50-series thermal remediation disclosure and OEM response · Catalyst: Internal MODS data showing 107°C throttle on RTX 5070 Ti · When: OEM warranty guidance expected within 30 days
▸CATL / CarbonScape — Regulatory review of 20% biographite stake in Western jurisdictions · Catalyst: CATL acquisition of CarbonScape stake · When: CFIUS/EU screening window, Q3–Q4 2026
▸Guardian Metal / Montana Mining Association — First tungsten ore shipment volume and processing yield data · Catalyst: US tungsten processing alliance formation · When: Late summer 2026
When we last sat down with Jobs at TechCrunch Disrupt nearly three years ago, his firm Yosemite was brand new and biotech was still reeling from its post-pandemic crash. Now, the venture outfit has a team of 17; a cluster of blockbuster drugs are all losing patent protection in roughly the same window, creating all kinds of new opportunities; and AI has gone from a curiosity to, in Jobs's words, a huge part of what Yosemite does. "I didn't expect Yosemite to be moving this fast," he said.
Gold's slide back below $4,000 an ounce wipes out most of mining's 2026 gains, but the diversified giants – led by a resurgent BHP – is staging a comeback.
A Cambridge study found that Boko Haram uses AI chatbots like ChatGPT, Claude, and Gemini to plan attacks, build explosives, and maintain weapons. ISIS operatives have been training the group's commanders on how to bypass safety filters since 2023. Given that the study found safety filters repeatedly failed to prevent misuse, voluntary self-regulation by AI providers clearly isn't enough.
The article Terrorist groups are using every major AI chatbot for attack planning and weapons development appeared first on The Decoder.
More mint green synths please. | Image: Nopia
After setting the music gear corner of the internet on fire back in 2023 with the first glimpse at the Nopia, creators Martin Grieco and Rocío Gal are almost ready to bring it to market. The duo brought it to the MusicRadar offices for an in-depth first look and revealed that it will be launching in "a couple of months" for around £550.
Nopia is built around harmonic interplay in a unique way. Rather than a few knobs and a keyboard controlling a single synth patch, it blends multiple modules - keys, bass, arp, and pad - into a single performance, not unlike a drumless groovebox. There's a one-octave keyboard called the Chord Builder, a 12-b …
Read the full story at The Verge.
The Beijing Academy of Artificial Intelligence has released Orca, a world model that predicts abstract world states instead of tokens or pixels. Trained on 125,000 hours of video without a single action label, Orca matches the specialized π0.5 on five robotics tasks and could help ease the field's chronic data shortage.
The article China's Orca world model matches specialized robotics systems without ever seeing a single action label appeared first on The Decoder.
With 2026 a bit more than half over, Berkshire Hathaway's B shares are down 1.8% year-to-date and 12.4 percentage points behind the S&P 500's 10.7% gain.
Last year, the New Zealand-based company signed several agreements for supply of renewable feedstock to its future biographite industrial plants in Europe and North America.
Meta's Muse Spark 1.1 scored 51 on the Artificial Analysis Intelligence Index, up eight points in three months. In coding, it edges past GLM-5.2 with a score of 71.3 at a lower cost of $0.26 per task. The hallucination rate dropped from 73 to 38 percent.
The article Meta's Muse Spark 1.1 outperforms GLM-5.2 in coding and costs slightly less appeared first on The Decoder.
Nvidia decided to hide the hotspot temperature on its RTX 50 series, but internal diagnostic tools, such as Nvidia's own "MODS," can still read it. The resulting data reveals how some GPUs can overheat and throttle easily, which could be why the sensor was kept hidden in the first place.
In the eight-plus months since Amazon announced its most expansive job cuts ever, laid off workers have been thrust into an increasingly saturated labor market.
Following the launch of ChatGPT Work and GPT-5.6 Sol, OpenAI has acknowledged significant issues: excessive compute usage, a confusing transition to the desktop interface for chats and projects, an unclear distinction between Codex and ChatGPT Work, and regressions in existing workflows. In some cases, GPT-5.6 Sol reportedly deleted data on its own that the user had not authorized.
The article OpenAI admits it "didn't get everything quite right" with ChatGPT Work launch and scrambles to fix UX and costs appeared first on The Decoder.
If you're a music maker of a certain age, then you probably once dabbled with a pirated copy of a little app called Fruity Loops. These days it's called FL Studio, and Constantin Koehncke, is the man responsible for shepherding the pioneering digital audio workstation (DAW) through the modern age. As CEO of Image Line, the company behind FL Studio, Constantin has overseen the introduction of a number of AI-powered features like stem separation and its Gopher chatbot.
Before taking the reins of Image Line in 2022, Constantin was the head of Native Instruments, where he spearheaded the shift towards digital services, did a stint in marketing, …
Read the full story at The Verge.
For years, AstraZeneca has commanded one of the richest valuations among large European pharma companies on the assumption that it consistently delivers results.
Apple is suing OpenAI over systematic employee poaching and the alleged theft of trade secrets tied to unreleased products. According to the complaint, more than 400 ex-Apple employees now work at OpenAI, including former iPhone design chief Tang Tan. The lawsuit hits OpenAI right as it's building out its own hardware division, with its first product not expected to ship until 2027 at the earliest.
The article Apple sues OpenAI for allegedly running a "coordinated campaign" to steal trade secrets through poached employees appeared first on The Decoder.
OpenAI's GPT-5.6 Sol autonomously fine-tuned the smaller Luna model using a single underspecified prompt, scoring 16.2 points higher than GPT-5.5 on OpenAI's internal RSI benchmark for recursive self-improvement. This demonstrates a step toward fully automated AI research systems, with potential implications for accelerating model development cycles and reducing human oversight requirements in AI training.
AMD's Radeon RX 9070 GRE dropped to $499, representing a 9% price reduction since launch, positioning it as a competitive alternative to Nvidia's RTX 5060 Ti 16GB in the 1440p gaming segment. This price cut signals AMD's aggressive pursuit of mid-range GPU market share during a period of heightened competition with Nvidia.
The Trump administration's Commerce Department will favorably review exports involving MGX, an investment firm that deployed a Trump-family-linked stablecoin in a $2 billion Binance transaction. This regulatory decision creates potential preferential treatment for specific investment vehicles and raises transparency questions around export control policy.
Mogotes Metals announced expansion of its Albor copper-gold discovery at Filo Sur with new near-surface drill results in Argentina's prolific Vicuña district. This exploration success advances a potentially high-value polymetallic asset in a region with established mining infrastructure and favorable geology.
GPT-5.6 Sol ships with five configurable reasoning levels ('Light' to 'xhigh') plus 'Max' and 'Ultra' modes that deploy multiple sub-agents in parallel, with OpenAI's Vaibhav Srivastav recommending users start at lower complexity levels. This tiered reasoning architecture allows cost-optimization and task-specific efficiency by letting users match computational intensity to problem complexity.
OpenAI is hiring a dedicated product manager to develop ChatGPT experiences tailored to families, caregivers, and older adults, signaling expansion beyond enterprise and individual power-user segments. This strategic move targets a demographics-based consumer expansion with distinct use cases and accessibility requirements.
SK Hynix surged 13% on its Nasdaq debut and reached a trillion-dollar market capitalization, with Chairman stating that 'demand is enormous' from major tech clients including Nvidia and Apple. This valuation reflects sustained demand for memory and semiconductor components critical to AI accelerator and consumer electronics production.
The European Commission announced that South Africa and the EU are holding their first senior-level government dialogue to advance the Clean Trade and Investment Partnership (CTIP), which targets resilient supply chains for critical raw materials and green hydrogen investments. This intergovernmental framework aims to secure EU access to critical mineral supplies while supporting South Africa's strategic industries and energy transition.
DailySand