Building AI for Lawyers - Jacob Lauritzen, CTO of Legora (opens latent.space)
Evals without traces, million-cell tabular reviews, and cracks in the billable hour as a business model
468 stories across editions, newest first.
Evals without traces, million-cell tabular reviews, and cracks in the billable hour as a business model
Memo: Linux distribution vendor SUSE is laying off ~10% of its employees, saying “automation and AI are fundamentally shifting how software companies operate” — One in ten employees at SUSE are facing the chop due to “automation and AI,” according to a memo dispatched by management …
Record venture funding in 2026 has been increasingly concentrated in massive AI rounds, while seed- and early-stage deal counts are on the decline. Using Crunchbase data we provide a look at how third-quarter funding continued the year's dollars-up-but-deals-down trend.
Listen now | 🎙️ Figma growth PM Ezra Mechaber demos PM OS, a 31-skill repo used by 60+ PMs, and AutoXP, an agent that runs growth experiments with one human checkpoint
About 16.5k books mentioned by at least two HN commenters since 2007, each linked back to the comments; built with Claude after Karpathy's HackerTube.
Hex says it rebuilt its data analytics platform for AI agents; Peak XV led the $100M round at a $1.2B valuation.
Broccoli AI sells AI agents to plumbing, HVAC and electrical companies; the Series A and Series B were announced together, with a source putting the valuation at $300M.
Aramco contributes about $2.6B in cash in 2026 plus AI, compute and digital assets for 49% of Humain; Saudi Arabia's PIF keeps 51%.
Hyperbolic lets companies rent AI compute across about 30 neoclouds; Walden Catalyst, Intel CEO Lip-Bu Tan's fund, led the round and total funding is $50M.
Maxinsights has nearly 4,000 collectors wearing cameras to record everyday actions and turns the footage into robot training data; its founder says the company holds about 2 million hours of first-person data and ARR is around $250M.
Sweep speeds up customer migrations to ServiceNow's CRM; the price was undisclosed but is estimated in the hundreds of millions.
Rune's AI software helps the military plan supplies near the front line; it joins Missionforce, the defence business Salesforce launched a year ago.
OpenAI’s latest funding round doesn't seem to be living up to expectations.
Google's prototype data centre satellite ran a Gemma 4 model in orbit and sent back a haiku, Google's Travis Beals wrote on Friday.
Axios reports OpenAI will push mandatory "speed limits" for powerful AI and rebrand its data centers as "AI compute campuses" after Tuesday's midterms.
A credit card for AI spend: faster inference, better GPU access and 4% back on AI.
Google's data center energy head says space data centers cost up to 100x more than Earth-based ones, and US power supply is sufficient but permits are slow — Data centers in space could cost up to 100 times as much as those on Earth, Google's head of data center energy said this week …
I’m a civil engineer and own a small municipal engineering firm. I also do forensic engineering, including expert work for legal cases. I use Claude, and it’s honestly shocking how powerful it is for this kind of work.
Apple's new CEO defended the company's current position in the AI landscape.
A Hugging Face repo claims to hold an internal Anthropic Mythos-class training checkpoint; a follow-up post says the shards it downloaded are fake.
The $100-a-month Max 5x plan is now Max and the $200 Max 20x plan is Ultra, which loses its five-hour usage limit; prices stay the same.
Hector Serrano on building Wispr Flow's paid-UGC machine and hiring young talent over polished resumes.
A federal judge in Mississippi refused to dismiss the NAACP's Clean Air Act suit over the gas turbines powering xAI's Colossus 2 data centre.
Shyam Sankar told Ben Thompson that Palantir was caught in Anthropic's clash with the US Department of War and does not regret working with Anthropic.
For those who have AI features included in a flat subscription: how do you stop one user from burning through your margin? What I'm curious about: - Do you set a hard cap per user (tokens, requests, $)? What happens when they hit it? - Rate limits, credits, pricing tiers, or something else?
The artificial intelligence buildout is making an enormous bet on what businesses will eventually be willing to pay for intelligence. In our research this summer on forecasting the AI bubble, we argued that a key indicator to watch is productive utilization by enterprise customers.
Game developers report Claude Code refusing work on projects that recreate existing games since late October, some of which they say they have a right to build.
The claim is that a team of five Qwen3.8-27B models orchestrated by Memdoor beats Fable 5 on LiveCodeBench Hard; it is the developer's own benchmark.
Few people shop through outside AI agents yet, because the experience is not good enough, Amazon chief executive Andy Jassy told analysts on Thursday. He expects Amazon’s own assistant, Alexa for Shopping, to be the main agent people use to buy from the store.
John Ternus says Apple doesn't “want to ship AI for AI's sake” and that Tim Cook is “available to help wherever I ask” on policy; Ternus has yet to meet Trump — Apple doesn't need fixing, and it doesn't want drama. Its new leader is a “product guy” who bleeds six colors.
Worked on an AI product where the co-founder pulled the plug because token costs and API changes kept shifting the unit economics every month. Now reading that enterprise AI runs $9-19M/year with no proven ROI. Anyone here actually showing positive unit economics on an AI product after 12+ months?
Sources: Meta built a prototype humanoid robot with Chinese firm Fourier in three months and is building a second; Meta says it has no plans to sell these units — Meta Platforms has built a prototype humanoid with Chinese robotics company Fourier and is already working on a second version …
The AI logistics startup would bolster Salesforce's Missionforce defense platform.
Researchers are paid $5K for reporting a bug to OpenAI that would let attackers run code in logged-in sessions of all who opened the hacker's ChatGPT share link — A malicious ChatGPT share link could run an attacker's code in the logged-in session of anyone who opened it …
Google, Microsoft, OpenAI and Anthropic will honour one shared robots.txt rule that keeps a site's content out of generative AI training.
Amazon's free cash flow fell to minus $42.4bn over the 12 months to September, mostly because of AI spending.