When serializing data to JSON, CSV or when logging, we convert numbers to strings. Floating-point numbers are stored in binary, but we need them as decimal strings. The first formally published algorithm is Steele and White’s Dragon schemes (specifically Dragin2) in 1990. Since then, faster methods have emerged: Grisu3, Ryū, Schubfach, Grisu-Exact, and Dragonbox. In C++17, we have a standard function called std::to_chars for this purpose. A common objective is to generate the shortest strings while still being about to uniquely identified the original number.

We recently published Converting Binary Floating-Point Numbers to Shortest Decimal Strings. We examine the full conversion, from the floating-point number to the string. In practice, the conversion implies two steps: we take the number and compute the significant and the power of 10 (step 1) and then we generate the string (step 2). E.g., for the number pi, you might need to compute 31415927 and -7 (step 1) before generating the string 3.1415927. The string generation requires placing the dot at the right location and switching to the exponential notation when needed. The generation of the string is relatively cheap and was probably a negligible cost for older schemes, but as the software got faster, it is now a more important component (using 20% to 35% of the time).

The results vary quite a bit depending on the numbers being converted. But we find that the two implementations tend to do best: Dragonbox by Jeon and Schubfach by Giulietti. The Ryū implementation by Adams is close behind or just as fast. All of these techniques are about 10 times faster than the original Dragon 4 from 1990. A tenfold performance gain in performance over three decades is equivalent to a gain of about 8% per year, entirely due to better implementations and algorithms.

Efficient algorithms use between 200 and 350 instructions for each string generated. We find that the standard function std::to_chars under Linux uses slightly more instructions than needed (up to nearly 2 times too many). So there is room to improve common implementations. Using the popular C++ library fmt is slightly less efficient.

A fun fact is that we found that that none of the available functions generate the shortest possible string. The std::to_chars C++ function renders the number 0.00011 as 0.00011 (7 characters), while the shorter scientific form 1.1e-4 would do. But, by convention, when switching to the scientific notation, it is required to pad the exponent to two digits (so 1.1e-04). Beyond this technicality, we found that no implementation always generate the shortest string.

All our code, datasets, and raw results are open-source. The benchmarking suite is at https://github.com/fastfloat/float_serialization_benchmark, test data at https://github.com/fastfloat/float-data.

Reference: Converting Binary Floating-Point Numbers to Shortest
Decimal Strings: An Experimental Review, Software: Practice and Experience (to appear)

Some good news: an important optimisation has rolled out for Excel PivotTables connected to Power BI semantic models! Back in 2019 I wrote about a very common problem affecting the MDX generated by PivotTables connected to Analysis Services where, when subtotals are turned off and grand totals are turned on, the query nevertheless returns the subtotal values. This led to extremely slow, expensive MDX queries being run and a lot of complaints. The nice people on the Excel team have now fixed this problem and PivotTables connected to Power BI semantic models generate MDX queries that only return the values needed by the PivotTable.

Here’s an example of a PivotTable connected to a published Power BI semantic model:

Note that the subtotals have been turned off but the grand totals are displayed – this is important. Here’s the MDX query generated for this PivotTable:

SELECT NON EMPTY 
{ /* GTOPT-BEGIN CSECTIONS=2 */ 
 /* GTOPT-SECT-BEGIN-1 Desc:GrandTotal */ 
{([Property Transactions].[New].[All],[Property Type].[Property Type Name].[All])}
 /* GTOPT-SECT-END-1 */ 
,
 /* GTOPT-SECT-BEGIN-2 Desc:Detailed */ 
{Hierarchize(CrossJoin({[Property Transactions].[New].[New].AllMembers}, 
{([Property Type].[Property Type Name].[Property Type Name].AllMembers)}))}
 /* GTOPT-SECT-END-2 */ 
} /* GTOPT-END */ 
 DIMENSION PROPERTIES PARENT_UNIQUE_NAME,HIERARCHY_UNIQUE_NAME ON COLUMNS  
 FROM [Model] 
 WHERE ([Measures].[Count Of Sales]) 
 CELL PROPERTIES VALUE, FORMAT_STRING, LANGUAGE, BACK_COLOR, FORE_COLOR, FONT_FLAGS

And here’s what this query returns:

There are 11 values displayed in the PivotTable and the MDX query returns 11 values. It’s what you’d expect but as I said, up to now, Excel would have generated an MDX query that returned 13 values – a query that also requested the subtotal values that aren’t displayed.

This optimisation should now be rolled out to 100% of Excel users. You can tell if you are using the new query pattern by looking for comments in the MDX code with the text “GTOPT” in – they’re easy to spot in the query shown above. Right now the optimisation only happens for PivotTables connected to Power BI semantic models but I’ve been told that in future it should also happen for PivotTables connected to Azure Analysis Services and SSAS; this is because some server-side optimisations are necessary to make the new MDX perform as well as possible.

You might be thinking that, despite my excitement, this is a very niche scenario but I assure you it’s not: Excel users very frequently create PivotTables that are formatted to look like tables, and having subtotals turned off and grand totals turned on is a key part of this. The more fields that are put on rows the more subtotals there are to calculate and the more the overhead increases; it’s not uncommon to find situations where the number of subtotal values is much greater than the number of values actually displayed in the PivotTable.

This doesn’t solve all the performance problems associated with PivotTables and Power BI though and more work is planned for the future.

[Thanks to Yaakov Ben Noon for driving this work!]

After I published my “velocity is the new authority” essay, a reader and dear friend emailed and asked if my framework explained the fraught relationship between media and the technology ecosystem it covers. My one-word reply was “Absolutely.” And here’s why.

In “comms” across tech, startups, and the larger ecosystem, little seems to matter anymore. It’s hard to pin down anythingconcrete or meaningful. Everything is noise and nothing can be heard.

Most founders (and technology companies) treat comms like a checklist: have something new, draft a story, chase press,issue a release, share it on LinkedIn and X, maybe do a podcast tour. Others skip intermediaries and go straight to their audience. That worked when media was the gatekeeper and friction was high. Building your audience and owning your channels made sense. Today, anyone can go direct, and the noise has exploded. Rising above it gets harder every day.

Velocity is replacing authority as the organizing principle for information. You see it in the broken relationship between media and the technology ecosystem. The YouTube tech review I wrote about earlier highlights the core issue: access.

---

---

Access journalism is as old as media itself. These days though, it is table stakes. Not just in tech, but across all media. I didn’t play the access game, so I was not a preferred “leaker” for many technology companies. I got my stories the old-fashioned way. By being an annoying scoop hunter. It didn’t pan out often, but occasionally I got the big one. Like Microsoft buying Skype. It was a hard game to play. It’s even harder now.

A few months ago, I naively asked the founder of a purportedly disruptive company for an interview. I thought it would be a great interview, as he was doing something new and interesting and worth digging into. Obviously, he didn’t respond. He didn’t have to. Maybe he was worried that an old-school reporter like me would do my homework and ask basic, but tough questions. The kind of questions that used to be standard fare, but now they get you ignored.

A hundred years ago, tabloids printed whatever they were told, just to sell rags. Today it’s videos and podcasts. Don’t get me started on the “technology” ones, especially those creaming over “AI.”

Podcasters need guests. To build an audience they chase boldface names. Access depends on avoiding tough questions. Press them with unflattering ones and you get a reputation. No surprise we now have podcasters acting as mouthpieces for tech companies. The “interview” becomes a promotional vehicle. The host lobs softballs and nods along with whatever narrative the guest is spinning. Both lapse into MBA-speak. It’s an Oscar-worthy performance. The line between journalism and marketing blurs.

I don’t think this is an ethical failure; it’s structural. When your business model depends on downloads and views, you do what you must: keep cozy relationships with a small circle of founders, venture capitalists, and executives you’re supposed to cover critically.

AI podcasts are a perfect example. You hear the same talking heads saying the same things. They’re half-bullish, half-skeptical. One side sees transformation, the other disaster. As with any technology, the truth sits somewhere in the middle. What you need is contextual skepticism, technical scrutiny, and analytic rigor. That mix doesn’t exist because the system won’t support it.

---

---

There are fewer than a dozen journalists I can name-check those who don’t disappoint. Nilay Patel of The Verge for example. There are some veterans, but only a handful are newcomers. Good technology coverage needs more.

While building GigaOm, I knew we’d drown in pitch meetings without standards. I wrote a one-page guide for my writers covering startups. Beyond taking the time to understand their subject, I asked them to answer four questions before writing about a founder or a startup.

Begin with the founders. Look past LinkedIn summaries to their actual histories. Have they built and shipped real products? Do they understand the market they claim to disrupt? Are they domain experts or opportunists chasing the latest trend? And lastly, as a form of validation, who are their financial backers?

Founders were interesting only if you had enough context about the technology landscape to distinguish the novel from the incremental. You had to know whether their insight was real or just a repackaging of existing ideas with better marketing.

Investors mattered. With every deal, a firm put its reputation on the line. Their involvement signaled intent. If Sequoia or Kleiner Perkins backed a company, it meant their partners had done due diligence. The amount of capital and the stage said a lot about conviction and risk tolerance.

Assessing the market opportunity amid technology trends was the hardest part and required the most expertise. You needed to understand what was technically possible,, what the market was ready for, how the cost curves looked, andwhere the ecosystem dependencies were.

All this took time. I spent years watching markets evolve, covering hundreds of companies, and seeing patterns emerge. My understanding of technology grew through conversations with people who built it. Tony Li and Pradeep Sindhu walked me through the secrets of internet routing. From Andy Bechtolsheim and Desh Deshpande, I learned the ins and outs of switching. David Huber, the man behind Ciena, introduced me to optical networks. I learned the technology and, in turn, how to ask the right questions.

I was lucky. Back then, the internet didn’t move at neck-snapping speed.

In my VC role, I told founders to step back and think about why their stories mattered to anyone but themselves. I encouraged them to build real rather than transactional relationships with journalists, and to wait until they had something genuinely interesting to say instead of treating PR as a checklist.

It was good advice then, and it still is, in part, as a subset of a new approach. The system I described, where thoughtful storytelling and relationship-building mattered, required time on both sides.

---

---

Much like the rest of media, tech journalism, today relies on writers who lack the deep context to do the work well. They are young, overworked, pushed to churn, and don’t have time to spend weeks learning a market before covering it. You can’t learn on the job while filing three stories a day or recording several podcast episodes a month.

As I argued in my velocity piece, we now operate on vibes. When a firm like Andreessen Horowitz invests in everything at every stage, what does that signal? A16z’s backing no longer means what Sequoia’s meant in 2005, and a TechCrunch launch no longer means what it did in 2008. The technology ecosystem is noise. Its media outlets are just the outward expression of that noise.

I just read a Bloomberg piece about a startup called Upscale, which claimed it would take on Cisco and Broadcom with AI networking gear. It has raised a truckload of money and is valued in the billions, yet the story never told me the mostbasic thing: what it actually does.

The Wall Street Journal, the bastion of good financial journalism, recently spent many inches on the peccadillos of one of the founders of Thinking Machines. Thinking Machines is an AI company founded by former OpenAI employees and is now valued at $12 billion. The WSJ and a handful of other major publications covered this drama as if it were a story for the Daily Mail or Vanity Fair, yet none of them explained to readers what Thinking Machines does or why it is worth $12 billion. What is its technological edge? You know, the basic financial and technological questions most readers need forcontext.

This isn’t the first time I’ve seen empty stories dominate. In the late 1990s, the dot-com boom launched a wave of new publications, staffed by new writers. They reported fluff: financings, concepts, hires, and parties. The hard part was covering the technology and the financial realities.

We built systems that reward acceleration, and we act surprised when everything feels rushed, shallow, a little manic. The algorithm doesn’t care if something is true. It cares if it moves. Nothing moves like titillation, gossip, and startup psychodrama.

**

To see what works in this environment, look at Sam Altman and OpenAI.

Sam is everywhere. Always in the media, always talking. What matters is that he’s in the conversation. People are talking about OpenAI, about him, about their vision for the future of AI. The objective is to make OpenAI synonymous with AI. Tighten the equation until, in people’s minds, OpenAI equals AI. That constant presence, that relentless visibility, lets him raise money at ungodly valuations. Everything else is tactics.

What he says doesn’t have to mean anything; it’s no surprise he’s often self-contradictory.

Take advertising as an example. In a talk at Harvard in October 2024, Sam said, “The idea of intermingling ads and AI is uniquely unsettling to me. I think ads were important to give the early internet a business model, but they do sort of somewhat fundamentally misalign a user’s incentives with the company providing the service. I view ads as a last-resortbusiness model.”

By October 2025 his tune had changed. In January 2026, ChatGPT rolled out ads. “It is clear to us that a lot of people want to use a lot of AI and don’t want to pay, so we are hopeful a business model like this can work.”

In fifteen months, he went from “uniquely unsettling” to “hopeful that a business model like this can work,” from “I hate ads” to ads running in ChatGPT. Anyone can connect the dots, but nobody does because the system doesn’t reward it.

The system rewards the next Sam interview. It seeks the sound bite that gets attention, gets views, picks up speed.Everyone covers Sam’s latest thoughts on AI, the future, and whatever else he wants to talk about that day.

You hardly see anybody connecting the dots and writing, “Here’s what Sam Altman said about advertising six months ago, here’s what he’s saying now, and here’s what that tells us about OpenAI’s actual business model versus its stated principles.” That story doesn’t get written because it requires someone to remember what was said six months ago, compare it to what is being said today, analyze the gap between rhetoric and reality, and ask why the gap exists and what it means for the company’s actual strategy.

Don’t hate the player, play the game. Like I said, velocity trumps everything.

February 1, 2026, San Francisco

Velocity Is the New Authority. Here’s Why

Posted by Phalene Gowling, Product Manager, Google Play

To build a thriving business on Google Play, you need more than just data  – you need a clear path to action. Today, we’re announcing a suite of upgrades to the Google Play Console and beyond, giving you greater visibility into your financial performance and specific, data-backed steps to improve it.

From new, actionable recommendations to more granular sales reporting, here’s how we’re helping you maximize your ROI.

New: Monetization insights and recommendations
Launch Status: Rolling out today

The Monetize with Play overview page is designed to be your ultimate command center. Today, we are upgrading it with a new dynamic insights section designed to give you a clearer view of your revenue drivers.

• Optimize conversion: Track your new Cart Conversion Rate.

• Reduce churn: Track cancelled subscriptions over time.

• Optimize pricing: Monitor your Average Revenue Per Paying User (ARPPU).

• Increase buyer reach: Analyze how much of your engaged audience convert to buyers.

But we aren’t just showing you the data – we’re helping you act on it. Starting today, Play Console will surface customized, actionable recommendations. If there are relevant opportunities – for example, a high churn rate – we will suggest specific, high-impact steps to help you reach your next monetization goal. Recommendations include effort levels and estimated ROI (where available), helping you prioritize your roadmap based on actual business value. Learn more.

We recently rolled out new Sales Channel data in your financial reporting. This allows you to attribute revenue to specific surfaces - including your app, the Play Store, and platforms like Google Play Games on PC. 

For native-PC game developers and media & entertainment subscription businesses alike, this granularity allows you to calculate the precise ROI of your cross-platform investments and understand exactly which channels are driving your growth. Learn more.

The Orders API provides programmatic access to one-time and recurring order transaction details. If you haven't integrated it yet, this API allows you to ingest real-time data directly into your internal dashboards for faster reconciliation and improved customer support.

Level Infinite (Tencent) says the API  “works so well that we want every app to use it."

Continuous improvements towards objective-led reporting 

You’ve told us that the biggest challenge isn't just accessing data, but connecting the dots across different metrics to see the full picture. We’re enhancing reporting that goes beyond data dumps to provide straightforward, actionable insights that help you reach business objectives faster.

Our goal is to create a more cohesive product experience centered around your objectives. By shifting from static reporting to dynamic, goal-orientated tools, we’re making it easier to track and optimize for revenue, conversion rates, and churn. These updates are just the beginning of a transformation designed to help you turn data into measurable growth.