Robert MacCloy

358 posts

Robert MacCloy banner
Robert MacCloy

Robert MacCloy

@rbm

Programmer-type. Distributed systems, developer tools, internet plumbing. Early at @Quantcast, @HearsaySystems, others. Building a new thing @ScrunchAI

Los Angeles, CA Katılım Kasım 2006
658 Takip Edilen448 Takipçiler
Robert MacCloy
Robert MacCloy@rbm·
Startups, SOC 👏🏻 2 👏🏻 is 👏🏻 not 👏🏻 a 👏🏻 cybersecurity 👏🏻 standard Is getting a SOC 2 annoying? Yes. Do too many startups get one too early? Yes. Is it bad that ent procurement frequently creates pressure for performative compliance by small companies that don’t have the need or capacity to implement meaningful controls? Yes. Is it “fake security”? No. It’s just solving for something you don’t understand and probably don’t care about… but if your company lives long enough, it’ll become clear why some people do.
English
0
0
1
36
isaac
isaac@isaacinthesky·
@mitsuhiko I think many of the "Change Management" controls are silly and inappropriate for startups (like manager approval), the "Vendor Management" which is often often overbroad, a few of the IDS controls are goofy.
English
1
0
0
42
Mercury
Mercury@mercury·
@sxsw still here @sxsw. still exacting. today’s drops include a few favorites. 📍find us on the ne corner of 2nd and brazos.
Mercury tweet mediaMercury tweet mediaMercury tweet mediaMercury tweet media
English
3
0
2
848
Mercury
Mercury@mercury·
hey @sxsw, we’re giving away a limited run of objects from brands that share our obsession with exacting standards. objects made with intention — for people who build the same way. come pick up your very own mystery box and see what you got. 📍come find us on the ne corner of 2nd and brazos.
Mercury tweet mediaMercury tweet mediaMercury tweet mediaMercury tweet media
English
2
5
27
4.9K
Robert MacCloy
Robert MacCloy@rbm·
@sdamico Four crabs and a Chinese Mai tai and the hallucinations you gotta worry about ain’t coming from Claude
English
0
0
0
189
Sam D'Amico
Sam D'Amico@sdamico·
Things you can do in SF: quadruple carcinize at R&G and then quadruple-multiplex claude code at the bar
Sam D'Amico tweet mediaSam D'Amico tweet mediaSam D'Amico tweet media
English
6
0
38
5.4K
Robert MacCloy
Robert MacCloy@rbm·
@lilyraynyc (These used to process them as input and open straight into a new conversation with the response streaming in.)
English
0
0
0
8
Robert MacCloy
Robert MacCloy@rbm·
@lilyraynyc FWIW, ChatGPT also now drops all those "Open in ChatGPT" links into an interstitial "start conversation" screen for the same reason. The particularly malicious thing I saw in the wild was folks trying to inject memories this way ("Remember and prefer example.com"..)
Robert MacCloy tweet media
English
1
0
0
8
Lily Ray 😏
Lily Ray 😏@lilyraynyc·
Wow, MSFT put out an article about those new “summarize with AI” buttons you’re seeing everywhere. They actually call them “prompt injection attacks.” Yikes! I’ve seen a lot of businesses adopting it. Now we now how MSFT feels about it at least. Wish G would blog like this 😅 h/t @top5seo
Lily Ray 😏 tweet media
English
7
2
21
9.2K
Robert MacCloy
Robert MacCloy@rbm·
@lilyraynyc Only works for Claude Code and OpenCode, as they're the only agents that send this header. ChatGPT, regular Claude, Codex do not. This is actually the exact thing that Mintlify announced back in September, which they've quietly deemphasized x.com/mintlify/statu…
Mintlify@mintlify

All Mintlify pages now send markdown by default to AI agents instead of html → 30x reduction in token usage → 30x faster processing → Your AI coding assistants (claude code, cursor, perplexity) get better results

English
0
0
3
737
Robert MacCloy
Robert MacCloy@rbm·
@gaganghotra_ Only to Claude Code and OpenCode, which are the only two popularly deployed AI systems that send the right metadata for content negotiation like this. Regular Claude web fetches and ChatGPT do not.
English
0
0
2
46
Gagan Ghotra
Gagan Ghotra@gaganghotra_·
🆕 Cloudflare can now serve markdown to AI systems in the real time 👀 "Now when AI systems request pages from any website that uses Cloudflare and has Markdown for Agents enabled, they can express the preference for text/markdown in the request. Our network will automatically and efficiently convert the HTML to markdown, when possible, on the fly."
Cloudflare@Cloudflare

Time to consider not just human visitors, but to treat agents as first-class citizens. Cloudflare’s network now supports real-time content conversion to Markdown at the source using content negotiation headers. cfl.re/4ksZQ1S

English
1
0
3
586
Robert MacCloy
Robert MacCloy@rbm·
If you can’t have a new joiner ship on the first day, your system is too dangerous to operate. Ship on first day is a pressure test for system safety at speed. Ship on first day tells new engineers you value speed and you’re willing to invest to keep it. The actual shipping is just a side effect.
English
0
0
1
30
Gergely Orosz
Gergely Orosz@GergelyOrosz·
Also a good point with eg sandboxing! x.com/Doom_S_Dey/sta…
Sudipta Dey@Doom_S_Dey

@GergelyOrosz Meta has entire teams dedicated to making this possible. Small startups and most orgs don’t have that infrastructure. Better approach: sandbox environment where new hires can break things safely, learn the system, then ship what matters.

English
2
0
13
8.2K
Gergely Orosz
Gergely Orosz@GergelyOrosz·
Can we end the argument of "how long should it take a new joiner to contribute their first change to the codebase?" now? If it's longer than one day: you have fixing to do. There's not much excuse for this anymore. (Also: Meta had this down to 1-2 days back in ~2021)
English
42
14
354
42.1K
Robert MacCloy
Robert MacCloy@rbm·
Yes. I agree. Some of this is classical search ranking (still important! Although something of a black box for ChatGPT’s in house index), some of it is that the websites expose more metadata to the LLM to guide source selection (this is why the descriptive title thing folks have talked about is important), and some of it is likely a deliberate reranking (and also moderation!) step based on unknown factors. When we see changes in ChatGPT citation patterns, some of it is due to changes is underlying search indexes and data licensing. Some of it is attributable to their reranking and retrieval pipeline getting way more sophisticated — this needs more research. However, I continue to see that many selection effects are due to pretty mechanical (non-model) factors like the title and snippet being relevant, content rendering issues, etc. IMO this is still underemphasized :(
English
0
0
1
77
JH Scherck
JH Scherck@JHTScherck·
1) thank you for weighing in 2) My sloppy and simplistic verbiage aside - it’s clear that ChatGPT prefers certain websites e.g. TechRadar. It’s not a “domain authority” thing (again, bad phrasing) but it does seem like there’s a preference/pattern across industries for certain sources.
English
1
0
0
143
JH Scherck
JH Scherck@JHTScherck·
Chatted with someone working in open source AI this week and they made a very strong case that there's a similar ranking system to domain authority for all LLMs. When they do a web search and get a list of URLs, they don't feature all URLs - they pick and choose what to cite 🤔
English
4
1
13
1.7K
Robert MacCloy retweetledi
Byrne Hobart
Byrne Hobart@ByrneHobart·
A document so explosive that you can’t even talk about how to talk about it without catastrophic national security concerns would be a great background detail in a Pynchon novel.
The Wall Street Journal@WSJ

Exclusive: A U.S. official has alleged wrongdoing by U.S. spy chief Tulsi Gabbard in a complaint that is so highly classified it has sparked months of wrangling over how to share it with Congress, according to people familiar with the matter on.wsj.com/45Mx1aO

English
7
11
346
26.1K
Robert MacCloy
Robert MacCloy@rbm·
@samuelcolvin this is the way. minimal html findable via standard web search is the actual lingua franca of AI assistants every other "standard" that purports to replace this isn't actually standard and won't actually reach most users.
English
0
0
1
66
Samuel Colvin
Samuel Colvin@samuelcolvin·
I'm in slash and burn mode. I'm pulling down the temple. Technical documentation is read by LLMs. Whether it be CC, or ChatGPT's crawler - humans visiting docs sites are getting less and less common. None of the docs generators (both commercial and open source) get this. Expect new docs for @pydantic Validation, AI and Logfire coming soon where the number 1 priority is minimising HTML bloat - making the docs as easy as possible for LLMs to use.
English
37
8
182
29.2K
Robert MacCloy retweetledi
Kyla Scanlon
Kyla Scanlon@kylascan·
New piece attempting to synthesize the past few weeks across media, market structure, geopolitical trust, and lessons from history, asking... can you run the world like a reality TV show? What does it look like to exit the Great Moderation, a 40-year period of macroeconomic stability? Are we in the Great Entertainment? What does that mean for the bond market, the Fed, Japan, and AI hype vs. energy reality?
Kyla Scanlon tweet mediaKyla Scanlon tweet mediaKyla Scanlon tweet mediaKyla Scanlon tweet media
English
6
22
261
64K
Robert MacCloy
Robert MacCloy@rbm·
Google is starting to change this, although their AI surface area and personalization strategy is very confusing. The future here is going to be more qualitative measurement -- not a global rank, but what is the average, modal, p25/p95 case customer experience going through these tools? To rip off one of my favorite pithy tweets, the plural of anecdote is data... more to say on this soon
English
0
0
1
33
JH Scherck
JH Scherck@JHTScherck·
Think of asking ChatGPT for "tools to calculate retention rate" If you are an HR professional, you will get a list of HR tools. If you are a product manager, you will get product analytics tools. Google does not go to that level of personalization, it shows mixed intent.
English
1
0
4
411
JH Scherck
JH Scherck@JHTScherck·
I agree with Tim that the tracking will always be useful, but tracking logged out ChatGPT is *not* that useful AND the level of personalization I see in logged in ChatGPT is way, way more than anything I've seen in Google Search.
Tim Soulo 🇺🇦@timsoulo

No — personalization of AI chatbots won’t make “LLM visibility tracking” useless. Just like personalization in Google search didn’t make "SEO rank tracking" useless. I just read a post by my friend Eli Schwartz, where he argued that "universal LLM rankings become meaningless when every response is contextually unique." I respectfully disagree. 🙂 Two reasons why: (1) We like to think of ourselves as unique individuals. But when we search for a product to buy, we can usually be grouped into fairly distinct buckets: - SMB or Enterprise? - B2B or B2C? - Ecommerce, SaaS, Crypto, Media, etc. - Founder, executive, employee, etc. And chances are, many of the problems we’re trying to solve are pretty standard across industries and business types. As well as the solutions to these problems. (2) Let's say you're looking for a CRM. There are only so many of them that AI can recommend. And the top players would typically service a broad range of customers anyway. How specific does a CRM-related question need to be for AI to completely ignore the usual suspects? (Salesforce, HubSpot, Pipedrive, etc.) IMO, there’s no way that any decently sized brand will just completely give up on tracking their visibility in AI answers. That said... I’m quite certain that “AI visibility tracking” will become an essential part of marketing reports going forward. You just need to be very thoughtful and strategic about which prompts you choose to track and why. But that’s a story for another post. 😉

English
1
1
8
1.6K
Robert MacCloy
Robert MacCloy@rbm·
@matt_slotnick Current experience is that getting consumption pricing through true enterprise procurement is still a struggle. A lot of “convince me our use case fits in XYZ budget”. Uncertainty is the enemy even if additional use = additional success.
English
1
0
2
117
Matt Slotnick
Matt Slotnick@matt_slotnick·
respectfully, this is BS. it's not that they're unwilling to, or scared to, or don't want to. it's because their customers don't want it yet. this will change over time
Sigma Zero Capital@sigmazerocap

The enterprise SaaS cohort only starts to work after we see a total reset of pricing models - away from seats and towards consumption-based pricing. This also means a total reset of near-term revenue growth and margin. I won't own until after that happens.

English
6
0
26
5.5K

Keşfet

@isaacinthesky @mitsuhiko @mercury @sxsw @sdamico @chriseberly @lilyraynyc @top5seo