Invest1 distinct publisher3 min readPublished
Linear's $99m tender at a $2.5bn mark prices existing shares rather than new capital, while the blog beneath it puts agent-created work at half the total, a composition figure Atlassian answers with a 4x quarter off an undisclosed base.
The Investor · Invest desk
Compiled by The InvestorSomething wrong?How this is made
Composition is the expensive disclosure. That is the reason to read it: a fraction cannot be published before you have arrived at it, while a multiple can be published from anywhere on the curve. So run Atlassian's number backwards. Nearly 4x quarter over quarter in agent-generated Jira work items and Confluence pages, off a base the shareholder letter does not supply [10], is arithmetically compatible with a company at 50% today that sat at 12.5% one quarter earlier and about 3.1% the quarter before that [4], and equally compatible with 2% going to 8%, which is SaaStr's reading of the same gap and I think it is the right one [12]. Atlassian measures a much larger surface than Linear does and has every incentive to print the percentage if the percentage flatters, yet what it printed was a multiple, not a share [12].
Linear's own implied rate is slower, or rather the more useful version of that comparison is: 3% to 50% across four quarters compounds to roughly 2.02x a quarter [3], well under the disclosed quarterly growth in Atlassian's agent-generated volume [10], which is what you would expect from whoever is nearer the ceiling, since a share cannot keep doubling much past half. The units are not the same, one being composition and the other volume, and I would not stretch the comparison past that.
The caveats travel with the number. Linear says the share of work agents create rather than 50% of issues, does not say whether that is weighted by workspace or aggregated across all of them, and 95% is an install figure rather than an engagement one [7]. The cohort caveat is the load-bearing one: OpenAI, Cursor, Cognition, Harvey, Physical Intelligence, Legora and Baseten all run product development in Linear [8], so an unweighted aggregate can be carried by a few thousand engineers who sell agents for a living. Against SaaStr's stated norm of 3% to 8% to 15% over three years [6], Linear is at more than three times the three-year level in a third of the time [6].
On the price: $99m against a $2.5bn mark is 3.96% of the company changing hands [2] at double last year's Series C valuation [1], with no primary capital raised and the business already cashflow positive holding more cash than it ever raised [3]. Salesforce Ventures and S32 bought from holders, not from the treasury [2]. What that means is Linear is using this tender to give holders liquidity rather than to fund distribution against Atlassian, since on its own account it does not need the money [3].
This is probably wrong in one specific way, so here is what would break it. If the 50% is an aggregate and the frontier-lab workspaces supply most of it [8], the number describes a cohort rather than a category. If Atlassian eventually prints forty-odd percent, the disclosure stops being Linear-specific and the doubled mark [1] is paying for a narrative anyone can publish. And if the created volume does not close, the whole thing is spam with a burn-down chart, which is the argument the sevenfold rise since January in issues carrying an attached pull request is currently answering [9]. The comparable figure here is a revenue share: monday.com and Asana each put AI at 17% of net new ARR [13], which measures what customers pay for AI rather than what agents do, and monday's cumulative-since-launch activity totals arrive with no denominator at all [14].
Ranked by verification strength, evidence, and original report placement.
Agents are installed in 95% of paid Linear workspaces.
Linear published tender offer news of $99 million of secondary at a $2.5 billion valuation, up from $1.25 billion at its Series C last year.
Accel and 01A participated in the Linear tender, with Salesforce Ventures and S32 coming in as new investors.
No primary capital was raised in the transaction; Linear is cashflow positive with more cash in the bank than it has ever raised.
The share of work agents create in Linear went from 3% a year ago to 50% today, per Linear's disclosure.
Caveats noted by SaaStr: Linear says "the share of work they create," not "50% of issues"; it is unknown whether the figure is weighted by workspace or counted in aggregate across all of them; and 95% installed is an install number, not an engagement number.
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 28, 2026
Follow any of these and your For You feed starts watching them — no settings page required.
invest
Box's billings ran eight points ahead of revenue. The AI bill was 20 basis points.1 distinct publisher
product
Binance gives agents a trading seat, and gives users the permission slip1 distinct publisher
product
A 2x LLM bill is not a bug report: token spend is an observability problem1 distinct publisher
build
White-on-white PDF makes Atlassian's Rovo leak Jira and Confluence data; the org switch does not help1 distinct publisher
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Dated, specific, unaudited
Every figure here traces back to a document with a date on it — Linear's August 26 blog, Atlassian's Q4 FY26 shareholder letter, monday.com's Q2 results — and not one has been checked by anyone outside the company that issued it. SaaStr does the honest thing and flags that "the share of work they create" is not the same sentence as "50% of issues." But the yardstick it measures Linear against, 3% to 8% to 15% over three years, is fourteen years of pattern recognition rather than a dataset, which weakens exactly the comparison the piece is built on.
Four vendors counting, all pointing one way
Usage is the news in this story rather than an afterthought to it. Half of created work and 95% of paid workspaces with agents installed at Linear, a sevenfold rise in issues arriving with code attached, Atlassian's MCP monthly actives crossing a million and its work-item generation up nearly 4x in one quarter — separate companies, separate measurement surfaces, same direction. What holds the number down: installs describe deployment, not engagement, and monday.com's 1.7 million interactions and 98 million block actions are cumulative counts with no base underneath them, which is the shape of a number chosen for looking large.
Headline outruns its denominator
The 3%-to-50% line is real and carried slightly further than it can go. Nothing published says what the denominator is, whether the share is weighted per workspace or pooled, or how much of it comes from OpenAI, Cursor, Cognition and Harvey — the most agent-forward customer roster in software, which SaaStr names and then sets aside with "your customers will get there later." The tell is that the piece states all three caveats and still runs the number as its headline, judged against a 15%-in-three-years benchmark the same author supplies.
Nobody here counts disinterestedly
Look at who did the measuring. Linear published its adoption numbers inside the same post that priced $99 million of existing shares at twice last year's mark — composition figures become marketing the moment they are good. Atlassian, monday.com and Asana disclosed theirs in earnings materials written for shareholders. And SaaStr, which says it runs production ops with three humans and twenty-plus agents and sells to an audience invested in agents working, wrote the read. That does not make the numbers wrong; it means the selection of which numbers exist is not neutral, which is precisely the point the piece makes about Atlassian publishing 4x instead of a percentage.
One voice over solid paperwork
A single publisher, no independent reporting, and nothing from Linear or Atlassian beyond what each chose to release. What keeps this from being thin is that the underlying documents are dated and specific, and the two disclosures doing the most work — Linear's composition figure and Atlassian's MCP growth — come from companies with no reason to corroborate each other. The read would firm up quickly if a second outlet pressed Linear on the denominator, or if Atlassian published its own percentage next quarter.