Chapter 3 of Pew's April 2025 report comparing US adults with AI experts on what AI will do over the next two decades. 56% of the experts surveyed expect AI's impact on the US to be positive, against 17% of the public; 35% of adults expect a negative impact, against 15% of experts. The gap is widest on work and money: 73% of experts think AI will positively affect how people do their jobs versus 23% of the public, and 69% versus 21% on the economy, with a 40-point gap on medical care (84% versus 44%). Experts and the public broadly agree on the risks to democracy and journalism: only 11% of experts and 9% of the public expect AI to help elections, while 61% of experts and 50% of the public expect harm. Gender splits are large among experts — 63% of male experts predict a positive impact versus 36% of female experts. Free to read, with methodology and appendix tables.
Why I recommend it: Use this when someone tells you 'the experts say AI will be fine at work' — the numbers show experts and the public are describing two different futures, and the widest gap of all is about jobs. Read it with two limits in mind: the fieldwork was in 2024 and published 3 April 2025, so it predates a lot of what has happened since, and Pew's 'AI experts' are people who published or presented at AI conferences, many of them employed by companies building AI, which is exactly the group the optimism gap belongs to.
The full report of MIT's Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training, published 13 August 2026 after five months of meetings and outreach. The committee — students, faculty from every school, and staff from the MIT Libraries and Teaching and Learning Lab — was charged with assessing current AI use, identifying teaching and assessment innovations, and proposing an AI use policy. It sets out eight guiding principles (be humble, be bold, put humanity front and centre, lean into learning, teach with intentionality, no one size fits all, augmentation not automation, think beyond the classroom) and three groups of recommendations. Free to read online and free to download, with appendices and an FAQ.
Why I recommend it: The most useful part is the principles section — 'augmentation not automation' and 'teach with intentionality' are phrases you can borrow directly when you have to argue an AI policy to a school, a manager or a client. Be straight about what it is, though: MIT examining MIT, written for a residential research university, so its recommendations do not transfer unchanged to a community college, a bootcamp or a workplace.
A newsletter for university students and early-career professionals on building a career deliberately: networking with low effort, building a portfolio when you are not an artist, reputation-building as a young professional, early-career growth rules, job hunting after graduation, and working in and around AI. Launched in early 2026. Every issue in the recent archive is free to everyone, and the whole archive is open.
Why I recommend it: Best for the first few years of a career, where the advice gap is widest — the pieces on networking and on building a portfolio without being a designer are the ones to start with. It is one writer's experience rather than researched guidance, so treat it as a practitioner's notes: take the habits, check any claim about hiring against a source that publishes its numbers.
Stanford's Center for Research on Foundation Models runs HELM as a living benchmark for language and multimodal models. Rather than one score, it reports many models across many scenarios on multiple metrics — accuracy, calibration, robustness, fairness, bias, toxicity and efficiency — and publishes the leaderboards alongside the raw model outputs (predictions and prompts) so you can check a claim yourself instead of taking a number on trust. Separate leaderboards cover areas such as classic HELM, instruction-following, medical, legal and safety. All results and analysis are free to browse on the site, no account.
Why I recommend it: The place to go when a vendor quotes you a benchmark figure. HELM's real value is that it shows the prompts and the model's actual answers, so you can see what the score measured. Be aware of what it is not: it is a snapshot of the model versions and dates CRFM ran, so check the run date before comparing anything to a model released since, and a model missing from a leaderboard usually means nobody ran it, not that it failed.
Allwork.Space's write-up of a four-day Reuters/Ipsos poll that closed on Sunday 20 September 2026: 73% of Americans worry AI companies have not gone far enough to prevent serious harm, 55% favour slowing AI development, 39% say AI is having a negative effect on society (up from 36% the month before, the highest since Reuters/Ipsos began asking in March), and only 11% call it positive. Most respondents said federal officials, not the companies, should set safety standards. Free to read, no paywall.
Why I recommend it: Useful when you need a number for how the public actually feels about AI at work rather than how vendors say it feels. Two honest limits: this is Allwork.Space reporting a Reuters/Ipsos poll, so read the original poll before quoting a figure in writing, and a poll measures opinion, not job losses — it tells you nothing about how many roles AI has actually replaced.
A free daily publication covering the future of work: hybrid and remote policy, return-to-office mandates, flexible workspace and coworking, workplace technology and AI, and the employment law and labour trends behind them. Articles read in full with no subscription and no paywall.
Why I recommend it: Read this when you want to know what employers are actually doing about remote and hybrid work before you negotiate for it, or when you are weighing an offer that calls itself flexible. It is written for the workspace industry, so some coverage is trade news about office operators rather than advice for you; take the policy reporting and skip the property pieces.
