I imagine part of the performance issue is their encryption flow? Their search is sub par even with the on-device search enabled, but besides that i’ve been a happy customer for a few years now. Catch all domain, multiple domains, their cli lets you download all messages and I setup a RAG flow to better search.
Regarding outages, this is the first one I’ve actually noticed and affected me. Obviously not great, but maybe I’ve been deluded in seeing GitHub‘s fiasco of what acceptable means.
You can just make the tool calls restricted/scoped to whatever the calling account has access to (or in this case the repo)
That way even if the LLM broke out of the system prompt the worst case would be similar to a 404 or 401.
Why are we giving these processes super user access? No reason to have the executing loop/chat turns/tool calls be scoped to anything but the narrowest permissions.
If the agent truly needs data/permutations across different accounts or repos, treat the tool calls like any other API that needs to do that kind of work pre-LLM
I think this seems purposeful, as there's 2 opposing forces at play:
- Have a model that follows the users instructions
- Have a model that follows the system prompt instructions more
For the 'safety' argument (Re: Fable), they need these models to have basically a 2-tier instruction system, but given LLMs aren't great with actual Logic unless they program it out to test, this runs afoul and we get one or the other.
Feels like optimizing for either precision or recall, but can't have both
Accountants can be expensive, especially if your books are messy, or have poor accounting practices from the start.
Systems like quickbooks, hubspot, payment processors all have tiers where yes on paper they make it easy to properly setup good accounting practices, but you’ll spend an additional 500/month+ to get those features.
Hiring an accountant to clean up the books and do quarterly book keeping is equally as expensive if not more.
Especially for small service based businesses, where margins can be tight, revenue can fluctuate heavily MoM, committing an additional $6k+ per year just to keep books organized is non-trivial.
As an experiment, i gave all our finance data for 2025 to an agent, and it did quite well after spot checking. There may be a middle ground where users can do exports, verify with “real” software, and have agents handle contextual classification to considerably cut down costs
I’d slightly disagree, the profile of people who go missing is as important as a random chance there is a coincidence. Former military officers, high-level scientists. These individuals have training, money, and live in areas where this tends not to happen.
A disappearance of someone from the above background, vs someone who is say in midwest rural America or near areas where human trafficking crimes occur at a higher rate than normal, matters.
Further, their research/knowledge of sensitive government material also implies they likely have some form of overwatch or at least minimal monitoring for foreign agent threats from our government (or had in the past). Its not uncommon for high ranking military officials to have some form of training in counter surveillance tradecraft for this exact reason.
The odds these events are due to a foreign adversary given the multiple wars and geopolitical tensions are not negligible
Bit tinfoil hat, but outside of the phone becoming thinner and other design benefits with integrated batteries, I largely think the intelligence community has an aligned interest to keep phone batteries from being hot-swappable like pre-iphone days.
Given the drafts vs final version of the bill/policy, looks like the battery now must be more easily replaced vs true camera-like battery hot swap.
Even with an iPhone turned off, NSA can still listen:
Unfortunately much of China’s perverse tactics (they’ve done this in a wide array of industries) is to steal patented tech and trade secrets from companies outside China, subsidize the manufacturing and development etc, then sell their product at an artificially low price which kills the original company and good faith competitors as they cannot compete with the artificially lower prices.
Then once the dust settles they’re the only company which can handle large order sizes required for supply chains to build downstream products, and the world becomes further reliant on them.
Security concerns and national defense aside, a prime example pre-ban was Huawei layer 1 infrastructure products which far exceeded feature density, and cost effectiveness than competitors due to the subsidies. They’ve done similar tactics with solar panels.
This doesn’t imply China or their state sponsored companies never create novel tech, but there’s a hugely perverse system whose purpose is to illegally undercut competition overseas with no real recourse from the victim countries outside of total company bans. And even then, people find a way around the bans and the damage is already done to the original companies.
I wouldn’t be surprised if in the next 5-10 years the new and popular programming language is one built with the idea of optimizing how well LLM’s (or at that point world models) understand and can use it.
Right now LLMs are taking languages meant for humans to understand better via abstraction, what if the next language is designed for optimal LLM/world model understanding?
Or instead of an entirely new language, theres some form of compiling/transpiling from the model language to a human centric one like WASM for LLMs
I think it has a potential to raise a lot of the salaries of blue-collar positions in middle America, and then create demand for the trades over the next decade or so.
I find it unlikely that white collar positions will be switching drastically to blue collar unless they’re already on the fence about it or they’re not middle to high up in the white collar ladder (six figures+)
Despite how obtuse the current administration views are, this has been true for decades. The churn of new papers and hype around medicine/biotech is nothing new.
Says nothing about endemic reproducibility crisis of the social sciences.
Since student loans have been basically guaranteed (bankruptcies can’t erase student loan obligations, in an attempt to push rates lower) and tuition steeply rose, academic institutions’ ratio of administrators to students has skyrocketed to a bureaucratic mess, leading to a flywheel of higher education costs and incentivizing research for money’s sake over impact to the field.
Real impact would be reproducing notoriously iffy studies, but that doesn’t bring in the dollars.
Its just another layer of potential misdirection that BBC themselves, and many other news orgs, perpetuate. Im not surprised.
From first hand experience -> secondary sources -> journalist regurgitation -> editorial changes
This is just another layer. Doesn't make it right, but we could do the same analysis with articles that mainstream news publishes (and it has been done, GroundNews looks to be a productized version of this)
Its very interesting when I see people I know personally, or YouTubers with small audiences get even local news/newspaper coverage. If its something potentially damning, nearly all cases have pieces of misrepresentation that either go unaccounted for, or a revision months later after the reputational damage is done.
Many veterans see the same for war reporting, spins/details omitted or changed. Its just now BBC sees an existential threat with AI doing their job for them. Hopefully in a few years more accurately.
Defaulting to China stealing IP is a perfectly reasonable first step.
China is known for their countless theft of Europe and especially American IP, selling it for a quarter of the price, and destroying the original company nearly overnight.
Its so bad even NASA has begun to restrict hiring Chinese nationals (which is more national defense, however illegally killing American companies can be seen as a national defense threat as well)
Agree, I think the high cost of full time hires for entry level software jobs (total comp + onboarding + mentoring) vs investing in AI and seeing if that gap can be filled is a far less risky choice at the current economic state.
6-12 months in, the AI bet doesnt pay off, then just stop spending money in it. cancel/dont renew contracts and move some teams around.
For full time entry hires, we typically dont see meaningful positive productivity (their cost is less than what they produce) for 6-8 months. Additionally, entry level takes time away from senior folks reducing their productivity. And if you need to cut payroll cost, its far more complicated, and worse for morale than just cutting AI spend.
So given the above, plus economy seemingly pre-recession (or have been according to some leading indicators) seems best to wait or hire very cautiously for next 6-8 months at least.
The fact something is profitable (even vices) does not mean it requires regulations, unless the regulation in mind is direct or indirect cap on profit margins?
Ah yes, the secret design of pistols which go off at the slightest bump (its a lottery, only 1 in 1,000 chance!)
Revoke contracts, investigate the leadership who accepted the contract, and hold Sig criminally liable given they have internal documents from years ago acknowledging the fact.
I agree its a popular excuse, however unlike the blockchain craze there’s legitimate use cases of productivity improvements with AI.
And if you can (in some cases) substantially increase productivity, then logically you can reduce team size and be as productive with less.
With the right prompting, you can cut a copywriting team in half easily.
My business has one copywriter/strategist, who I’ve automated the writing part by collecting transcripts and brand guidelines from client meetings. Now she can focus on much higher quality edits, work with other parts of the strategy pipeline, and ultimately more clients than before.
I can easily imagine a corp with 100 junior copywriters quickly reducing headcount
Unfortunately i think many of those jobs can also be attributed to general economic health post low interest rates.
Companies now need to leave pre-revenue and turn a profit, or if you’re an established company you need to cut costs/increase margins from other economic headwinds (tariffs, inflation, gov policies etc)
A Junior dev (and most devs onboarding) will typically require 6-8 months to start being able to meaningfully contribute, then there’s a general oversight/mentorship for a few years after.
Yes they produce, however I think junior’s market salary plus the opportunity cost lost of the higher salaried mid and senior level in mentoring is a hard pill to swallow.
The team i work on is stretched very thin, and even after layoffs (which management agreed they went too far with) it’s pulling teeth to get another dev to build things companies are begging for and even willing to separately pay cash upfront for us to build
If you’re getting into the current job market as a junior, you’ll likely need to go heavy in the buzzword tech, accept a position from a smaller company that pays substantially less, then in 1-2 years job hop into a higher paying mid level role (not to say 1-2 years makes anyone mid level imo)
Regarding outages, this is the first one I’ve actually noticed and affected me. Obviously not great, but maybe I’ve been deluded in seeing GitHub‘s fiasco of what acceptable means.