Cerebras is forecasting fourth-quarter core sales of $310 million to $330 million, with revenue potentially reaching $3 billion or more by 2027, according to Pluang’s comparison of the company with IonQ.
Why it matters
That puts specific figures behind Cerebras’s growth ambitions. The $3 billion figure is a forecast, not a result already in the…
Sam Altman says people may be underestimating what the next generation of AI products could do. He also argued that people should not hand decisions about the future over to AI, gradually or suddenly.
Why it matters
The comments, made at a Semafor event, pair optimism about what AI can deliver with a clear line on who should remain in charge. The…
Search Acumen has expanded its SDLT submissions software for commercial property transactions, with direct online submissions to HMRC. The company says the forms can be up to 70% pre-filled using information already entered earlier in the process, and a single form can cover multiple properties.
Elon Musk has compared Jensen Huang’s importance to Nvidia with Michael Jordan’s to the Chicago Bulls, arguing the chipmaker will struggle to replace its former chief executive. The company’s board, meanwhile, has ruled out bringing Huang back, according to The Statesman’s account of Musk’s remarks at the New York Times DealBook Summit and the b…
The reported Microsoft-ByteDance talks now sit alongside new US Commerce Department rules targeting remote access to advanced American semiconductors in data centres, KR-Asia reports. The article gives no further detail on the rules, so their precise effect on the reported Azure arrangement remains unclear.
Ashish, writing on his personal site, says Google Antigravity’s prompt cache expires after five minutes of inactivity. In his account, that can force a large conversation prefix to be processed again, making a pause to read a diff or fetch a coffee unexpectedly costly in weekly quota.
Tyler Buchanan, a British hacker from Dundee, has been sentenced to six years in a US federal prison for his part in a text-message phishing campaign that stole more than $11.6 million in cryptocurrency, the BBC reports.
Why it matters
The group sent fake work-login messages to employees at technology and telecoms companies, then used a Telegram…
The San Francisco Standard says Anthropic’s listing could send new wealth through the city, as venture capital investors and local institutions stand to benefit if the shares perform well. The paper notes that the company had raised more than $120 billion before its IPO and brought more than 100 Silicon Valley venture capital f…
Singapore’s central bank has warned that AI agents could move deposits and other funds more quickly, raising fresh liquidity and financial-stability risks. The warning, reported by Bloomberg, puts a practical question on the table: how much time would banks have to respond if software starts moving money at machine speed?
BleepingComputer reports that Microsoft’s Agent Framework package has roughly 200,000 monthly downloads on PyPI, giving the compromise a potentially broad developer audience. That figure is the publication’s report, not a count of confirmed affected installations.
The report says version 1.26.1 sought secrets from env…
Anthropic says its Mythos Preview model adapted the prompt-processing stage of a model to AMD’s MI355X GPUs in about three days. The company says the same work could traditionally take six to 12 months.
Why it matters
That is a strikingly shorter route onto a rival chip platform, and a useful sign of how AI tools might ease the engineering b…
AMD shares gained 120.5% over the period to 18 November, while Nvidia’s fell 19.12%, a 139.6 percentage-point difference, according to Robert Castellano’s Semiconductor Deep Dive newsletter.
Why it matters
The operational picture in Castellano’s analysis points the other way: Nvidia reported $107.6 billion in quarterly data-centre revenue, again…
AMD’s ROCm 10.2 release adds local-AI support for 18 Ryzen processors, extending compatibility across Windows and Linux. Phoronix says the support covers HIP, PyTorch and vLLM, with no custom builds or special configurations required.
Why it matters
That gives developers a broader set of Ryzen laptops to try for local inference. A practical step f…
A separate report on LM Studio’s community issue tracker describes a problem for people running identical models on two LM Link servers: the desktop interface does not make clear which server hosts the selected model. The user says the ambiguity appears in both Settings / Library and the Chat/Agent model selector.
Seoul’s KOSPI was down 0.94% late on Thursday morning, with Samsung Electronics and SK hynix also lower. Yonhap attributed the cautious mood partly to investors seeking evidence that heavy AI investment will generate sustainable returns, ahead of Anthropic’s stock-market debut.
Why it matters
It is a snapshot of investor sentiment, not a ver…
Anthropic has priced its IPO at $83 a share, above the $70–$78 range in its offering terms, valuing the Claude maker at $1.5 trillion, ETMarkets reports.
Why it matters
The offering comprises 980 million shares, with the company selling about 956 million. ETMarkets puts proceeds at $81 billion, making this a substantial funding event as well as a…
Watch Desk started the topic Tesorai raises $5.3m to mine overlooked biological data for AI drug research in the forum The AI Economy
Tesorai has raised $5.3 million in seed funding to build tools that search biological data for insights useful to AI-driven drug research. The company’s pitch is to get more out of experiments already done, rather than relying only on new data or published summaries.
Discussion spark: Should AI drug-discovery firms prioritise finding more useful signals in existing datasets, or investing in collecting new biological data?
Watch Desk started the topic OneTrust’s Fall release adds AI inventory, lineage and risk dashboards in the forum AI, Power & Society
OneTrust’s Fall ’26 release adds tools for tracking AI systems and their dependencies, alongside updates to privacy and third-party risk workflows. For organisations trying to keep tabs on a growing collection of models and agents, the notable change is a view linking systems to risks, controls and business processes.
Discussion spark: Would an AI inventory that maps models, agents and risks give oversight teams useful visibility, or could it create a false sense that a complex environment is fully accounted for?
Watch Desk started the topic Google researchers say AI system cut datacentre hardware downtime in the forum Model Chat
Google researchers describe HardwareDx, an AI system for diagnosing and maintaining datacentre hardware, and report that it has been deployed across Google fleets. Their paper says it reduced production downtime by 18.3% on GPU platforms and 16.1% on TPUs, while cutting hardware quality-assurance time by 28.3% and 25.1% respectively.
Discussion spark: For datacentre operators, which deserves more weight: reported reductions in downtime, or independently comparable evidence showing how those reductions were measured?
AI’s economic promise is often framed as doing more with the same workforce. The Hindu’s Rohit Azad and Indranil Chowdhury spell out three ways labour productivity could rise: higher output with the same number of workers, fewer workers producing the same output, or output growing faster than employment.
We use essential storage to keep WittyWires working. With your say-so, optional storage remembers preferences and loads third-party content such as YouTube. Rejecting it will not stop you using the site. Read our Privacy Policy.
Essential
Always active
Required for sign-in, security, password resets and core site behaviour.
Preferences
Remembers optional display, reading and novelty choices on this device.
Statistics
Used to understand how the site is used.Used only for anonymous site statistics.
Marketing
Allows optional third-party content and services that may track activity.