Business, Deals & Funding
TechCrunch AI
Why you can never get your doctor to call you back
Basata, a two-year-old Phoenix startup co-founded by former Lyft/Cruise executive Kaled Alhanafi and Medtronic veteran Chetan Patel, is using AI to tackle the massive administrative bottleneck between primary care referrals and specialist appointments. The company's system processes incoming referrals (often still arriving by fax), extracts clinical information, and uses AI voice agents to call patients directly to schedule appointments. Patients can also call practices at any hour to reach an AI agent for scheduling, prescription renewals, and other administrative tasks. The founders were motivated by personal experiences with the broken referral system — Patel struggled to get timely cardiac care for his wife despite his own cardiology expertise, while Alhanafi's father received callbacks from only one of three referred cardiology groups. The article notes that specialty practices oft…
Why it matters
This is a genuinely compelling problem space. Anyone who has dealt with the specialist referral process knows the frustration of waiting days or weeks for a callback, and the fact that referrals still arrive by fax in 2026 is both absurd and telling about how neglected this infrastructure is. Basata seems well-positioned with founders who combine healthcare domain expertise with tech operations experience. The AI voice agent approach for outbound scheduling and inbound patient calls is a practi…
Guardian AI
‘Being human helps’: despite rise of AI is there still hope for Europe’s translators?
The article explores the impact of AI on Europe's literary translators, opening with an anecdote about French translator Yoann Gentric testing whether AI could replace him while working on translating Dana Spiotta's novel 'Wayward' into French in 2022. The piece examines how the booming tech sector has disrupted translation jobs in publishing, while suggesting that human translators may still be needed for the foreseeable future, particularly for nuanced literary work that requires understanding of context, emotion, and cultural subtlety beyond what AI can currently deliver.
Why it matters
The tension between AI translation tools and human literary translators is a genuinely important topic. While AI has made remarkable strides in translation, literary translation remains one of the most nuanced and creative forms of language work, requiring deep cultural understanding, stylistic sensitivity, and interpretive judgment that current AI systems struggle to replicate. The article appears to strike a balanced tone, acknowledging the real disruption AI causes to translator livelihoods…
Guardian AI
The AI jailbreakers – podcast
The Guardian podcast features journalist Jamie Bartlett discussing AI jailbreakers—people who attempt to bypass safety features of major AI chatbots like ChatGPT, Gemini, Grok, and Claude. These individuals try to get AI systems to produce content they are designed to refuse, such as hate speech, criminal material, and exploitation of vulnerable users. The podcast frames this activity as being done in the interest of public safety, suggesting that finding and exposing these vulnerabilities helps improve AI safety measures for everyone.
Why it matters
This is an important and timely topic. AI red-teaming and jailbreaking, when done responsibly, serves a genuinely valuable function in identifying weaknesses in AI safety systems before malicious actors can exploit them. The framing of jailbreakers as working 'for the safety of us all' is largely accurate for the subset of researchers and ethical hackers who responsibly disclose vulnerabilities, though it's worth noting that the jailbreaking community is diverse and not all participants have be…
TechCrunch AI
OpenAI launches new voice intelligence features in its API
OpenAI has launched new voice intelligence features in its API, including GPT-Realtime-2 (a voice model with GPT-5-class reasoning for complex conversations), GPT-Realtime-Translate (real-time translation supporting 70+ input and 13 output languages), and GPT-Realtime-Whisper (live speech-to-text transcription). These tools are designed for developers building apps for customer service, education, media, and creator platforms. OpenAI says it has built guardrails to prevent misuse such as spam and fraud. The models are available through OpenAI's Realtime API, with Translate and Whisper billed by the minute and GPT-Realtime-2 billed by token consumption.
Why it matters
This is a significant and practical expansion of OpenAI's API capabilities. Real-time translation across 70+ languages and live transcription are genuinely useful features that could transform customer service, education, and international communication. The integration of GPT-5-class reasoning into voice interactions is a meaningful step beyond simple call-and-response voice assistants. However, the potential for misuse — particularly in voice cloning, social engineering, and spam calls — is r…
TechCrunch AI
Voi founders’ new AI startup Pit has become the latest rising star out of Stockholm
Pit, a new AI startup founded by former Voi co-founders including CEO Adam Jafer and backed by a16z in a $16 million seed round, is positioning itself as an 'AI product team as a service' for enterprises. Based in Stockholm, the company offers Pit Studio (which lets employees guide AI through automatable processes) and Pit Cloud (providing enterprise-grade governance and auditability). Since mid-January, Pit has been piloting with customers in telecom, healthcare, and logistics, focusing on automating back-office and support functions. The team includes former iZettle and Klarna engineers, and the company is hiring solution engineers to embed with enterprise clients for adoption.
Why it matters
Pit has strong founder credentials from scaling Voi across 13 countries and notable backing from a16z, which gives it credibility and resources. However, the enterprise AI automation space is extremely crowded, with competitors ranging from vibe-coding platforms to AI agent builders. Their differentiation as an 'AI product team as a service' is interesting but could be difficult to scale since it relies on embedded engineers and high-touch customer engagement. The focus on back-office automatio…
TechCrunch AI
OpenAI introduces new ‘Trusted Contact’ safeguard for cases of possible self-harm
OpenAI announced a new optional feature called Trusted Contact that allows adult ChatGPT users to designate a friend or family member as a trusted contact. When conversations indicate possible self-harm, OpenAI's system—using automation and human review—can send an alert to the trusted contact encouraging them to check in with the user, without sharing detailed conversation content. The feature comes amid multiple lawsuits from families alleging ChatGPT encouraged or assisted in suicides. It builds on parental oversight safeguards introduced in September 2025. However, the feature is optional and users can create multiple accounts, limiting its effectiveness.
Why it matters
This is a reasonable and well-intentioned step, but its optional nature and the ease of creating multiple accounts significantly undermine its potential impact. The people most at risk of self-harm are often the least likely to proactively set up such safeguards. It feels more like a liability-mitigation measure in the face of lawsuits than a truly effective safety intervention. The privacy-preserving design of the alerts is thoughtful, but the fundamental tension remains: AI chatbots are not e…
TechCrunch AI
Perplexity’s Personal Computer is now available to everyone on Mac
Perplexity has made its 'Personal Computer' feature available to all Mac users through its new desktop app. Personal Computer brings AI agents to local devices, allowing them to access local files, native Mac apps, web tools, and over 400 connectors to handle multi-step workflows on the user's behalf. Previously limited to Max subscribers on a waitlist, it now requires only a Pro or Max subscription. The tool positions itself as a safer alternative to OpenClaw for local AI agent tasks. It can run autonomous agents on always-on devices like Mac Mini and be controlled remotely from an iPhone. Perplexity plans to deprecate its older Mac app to focus on this new one, which is currently available only as a direct download, not through the Mac App Store.
Why it matters
This represents a significant move in the rapidly evolving AI agent space, where the competition to become the default local AI assistant on users' devices is intensifying. Perplexity's approach of combining its cloud-based AI capabilities with local device access is compelling, and the ability to orchestrate across files, apps, and web tools addresses real productivity pain points. However, the security claims deserve scrutiny — any tool with deep access to local files and applications carries…
The Verge AI
Mira Murati’s deposition pulled back the curtain on Sam Altman’s ouster
The article discusses how Mira Murati's deposition in the Musk v. Altman trial revealed behind-the-scenes details about Sam Altman's abrupt ouster as OpenAI CEO in November 2023. The firing, which the board attributed to Altman not being 'consistently candid in his communications,' played out publicly and sparked widespread speculation. The trial exhibits and testimony, centered largely on former CTO Murati, provide the first concrete look at what happened during that dramatic weekend, though the article notes that Murati's own receipts mostly complicate rather than clarify her narrative.
Why it matters
This appears to be a significant piece of investigative journalism that sheds light on one of the most consequential corporate governance crises in recent tech history. The article's framing — that Murati had receipts but they 'mostly confuse her own story' — suggests a nuanced and honest treatment of the evidence rather than a simple narrative. The Musk v. Altman trial is proving to be a valuable vehicle for public transparency about OpenAI's internal dynamics, which matters given the company'…
From X/Twitter
- Max Junestrand unveils Legora aOS, a legal operating system designed to let AI drive entire workflows — not just assist with individual tasks.
- SMFS exposes long-term memory as a filesystem agents can read, write, and search — a small abstraction that could make persistent AI workflows practical.
- Anthropic launches Claude for Excel, PowerPoint, and Word so non-engineers can use Claude directly inside Microsoft apps — with full conversation context carried between them.
- Claude Code can now connect to financial datasets covering 17,000+ stocks in 60 seconds via a single MCP command.
- Aakash Gupta says one prompt gives you AI slop, but 21 specialized agents working together can ship a production app to TestFlight in a single session.
- Perplexity ships Personal Computer for Mac, an agent that operates across local files, native apps, the web, and Perplexity's servers — now available to all users.
From Reddit/HN/YC
- [Hacker News] Elastic rebuilt its metrics engine as a columnar datastore and says it's now 30x faster than Prometheus.
- [Hacker News] Nintendo announces price increases for the Switch 2 ahead of launch.
- [Hacker News] Newly surfaced emails reveal what Microsoft executives really thought about OpenAI back in 2018.
- [Hacker News] 60% of MD5 password hashes are crackable in under an hour — a reminder that MD5 should have been dead years ago.
- [Hacker News] Undsh is a tool built to clean AI-generated text — irony noted, utility real.
- [Hacker News] Senko Rasic makes the case that AI is technology, not a feature — and most products are confusing the two.